MTR05 B0 01 6R9

MITR E TECHNI CAL R EPORT

Test Procedures
for Verifying IAFIS Image
Quality Requirements for
Fingerprint Scanners and
Sponsor: DOJ/FBI
Printers, v1.5
Dept. No.: G036
Contract No.: J-FBI-07-164
Project No.: 1412FC23

Approved for Public Release;

Distribution Unlimited. 13-0939

Norman B. Nill
©2005-2016 The MITRE Corporation.
Margaret A. Lepley
All rights reserved. Chris F. Bas
Bedford, MA

October 2016
(blank page)

ii
 ABSTRACT

The Integrated Automated Fingerprint Identification System (IAFIS) is a national

fingerprint and criminal history system maintained by the Federal Bureau of Investigation
(FBI), Criminal Justice Information Services (CJIS) Division. The IAFIS provides
automated fingerprint and latent print search capabilities, digital image storage, and
transmission of fingerprints and query responses. The fingerprints and corresponding
criminal history information are submitted voluntarily by state, local, and federal law
enforcement agencies. IAFIS incorporates fingerprint capture devices (scanners) and
fingerprint printers, either of which can affect the image quality of fingerprint images that
are processed through the system. The FBI has therefore established an IAFIS Image
Quality Specification in order to define the quantitative image quality requirements for
these devices.

This document delineates the test procedures used to verify compliance of fingerprint
capture devices and printers with the IAFIS Image Quality Specification. It is intended to
support procurements and in-house development efforts throughout the criminal justice
community and, through the corresponding FBI product certification program, promote
and enhance interoperability between all IAFIS participants.

iii
(blank page)

iv
 DOCUMENT CHANGE HISTORY

Date Document Version1 Description of Change2

Number
October MTR v1.5 Minimum dimensions for FAP 50 increased.
2016 05B0016R9
February MTR v1.4 Update to sponsor contact [Eric M. Phillips] and
2016 05B0016R8 update to reference new 50 card test deck.
February MTR v1.4 This document, completed by M.A. Lepley, C. F.
2013 05B0016R7 Bas, and T.M. Nanez, modifies Section 7 related
to Stress Imagery captures and Environmental
Information.
May 2012 MTR v1.3 Update to sponsor contact [Eric M. Phillips]
05B0016R6
April 2012 MTR v1.3 This document adds additional procedures related
05B0016R5 to stress imagery and inked card comparisons.
June MTR v1.2 Update to sponsor contact [Frederick B. Jaco]
2011 05B0016R4
(dated
9/2010)
September MTR v1.2 Update to sponsor contact [John M. Manzo]
2010 05B0016R3
March MTR v1.2 Revision by C.F. Bas, added technical content to
2010 05B0016R2 Section 4 related to Fingerprint Printers.
September MTR v1.1 Revision completed by M.A. Lepley, based upon
2008 05B0016R1 revisions started by N.B. Nill.
Update to sponsor contact [B. Scott Swann]
April 2005 MTR N/A Test Procedures for Verifying IAFIS Scanner
05B0000016 Image Quality Requirements for Fingerprint
Scanners and Printers, by N. B. Nill.
An extensive revision of the 1994 MITRE
document: MP 94B0000039R1
November MP N/A Test Procedures for Verifying IAFIS Scanner
1994 94B0000039 Image Quality Requirements, by R.D. Forkert,
R1 G.T. Kearnan, N.B. Nill, and P.N. Topiwala.
In addition to the above MITRE documents, this
document formed the basis of the 1995 FBI
document: Test Procedures for Verifying IAFIS
Scanner Image Quality Requirements, FBI
document number CJIS-TD-0110, March 1995.

1 Version number is only increased upon a change in technical content.

2 More recent versions of the document supersede earlier versions.

v
(blank page)

vi
TABLE OF CONTENTS

SECTION PAGE

1 Introduction 1

2 Fingerprint Scanner 7
2.1 Linearity 7
2.1.1 Requirements 7
2.1.2 Background 7
2.1.3 Target 8
2.1.4 Test Procedures 8
2.1.5 Requirements Compliance 8
2.2 Resolution and Geometric Accuracy 9
2.2.1 Requirements 9
2.2.2 Background 10
2.2.3 Target 11
2.2.4 Test Procedures 13
2.2.4.1 Resolution and Across-Bar Geometric Accuracy 13
2.2.4.2 Along-Bar Geometric Accuracy 13
2.2.5 Requirements Compliance 14
2.3 Spatial Frequency Response 15
2.3.1 Requirements 15
2.3.2 Target 18
2.3.3 Test Procedures 19
2.3.4 Requirements Compliance 26
2.4 Signal to Noise Ratio 27
2.4.1 Requirements 27
2.4.2 Background 27
2.4.3 Target 27
2.4.4 Test Procedures 28
2.4.5 Requirements Compliance 29
2.5 Gray-Level Uniformity 31
2.5.1 Requirements 31
2.5.2 Target 31
2.5.3 Test Procedures 31
2.5.3.1 Preparation 31
2.5.3.2 Adjacent Row, Column Uniformity Test Procedure 31
2.5.3.3 Pixel to Pixel Uniformity Test Procedure 32
2.5.3.4 Small Area Uniformity Test Procedure 33
2.5.4 Requirements Compliance 33

v
 2.6 Fingerprint Image Quality 34
2.6.1 Requirements 34
2.6.2 Target 34
2.6.3 Test Procedures 36
2.6.3.1 Fingerprint Gray Range 36
2.6.3.2 Fingerprint Artifacts and Anomalies 38
2.6.3.3 Fingerprint Sharpness and Detail Rendition 39
2.6.4 Requirements Compliance 39

3 Identification Flats Scanner 41

3.1 Requirements 41
3.2 Background 41
3.3 Test Procedures 41
3.4 Requirements Compliance 42

4 Fingerprint Printer 43
4.1 Spatial Frequency Response 43
4.1.1 Requirements 43
4.1.2 Test Procedures 44
4.1.3 Requirements Compliance 45
4.2 Gray-Levels 45
4.2.1 Requirements 45
4.2.2 Test Procedures 45
4.2.3 Requirements Compliance 46
4.3 Dynamic Range 46
4.3.1 Requirements 46
4.3.2 Test Procedures 46
4.3.3 Requirements Compliance 46
4.4 Geometric Accuracy and Print Scale 48
4.4.1 Requirements 48
4.4.2 Test Procedures 48
4.4.3 Requirements Compliance 50
4.5 Noise 51
4.5.1 Requirements 51
4.5.2 Test Procedures 51
4.5.3 Requirements Compliance 51
4.6 Print Polarity and Color 52
4.6.1 Requirements 52
4.6.2 Requirements Compliance 52
4.7 Print Permanence 52
4.7.1 Requirements 52
4.7.2 Requirements Compliance 52

vi
 4.8 Print Stability 52
4.8.1 Requirements 52
4.8.2 Requirements Compliance 53
4.9 Hazardous Materials 53
4.9.1 Requirements 53
4.9.2 Requirements Compliance 53
4.10 Fingerprint Prints 53
4.10.1 Print Types Requirements 53
4.10.2 Requirements Compliance 54
4.11 Auxiliary Print Data 54
4.11.1 Labels 54
4.11.1.1 Requirements 54
4.11.1.2 Requirements Compliance 55
4.11.2 Bar Chart 55
4.11.2.1 Requirements 55
4.11.2.2 Requirements Compliance 56
4.11.3 Step Tablet 56
4.11.3.1 Requirements 56
4.11.3.2 Requirements Compliance 57
4.11.4 Finger Condition Codes 57
4.11.4.1 Requirements 57
4.11.4.2 Requirements Compliance 57
4.12 Fingerprint Quality 57
4.12.1 Requirements 57
4.12.2 Test Procedures 58
4.12.3 Requirements Compliance 58

5 Mobile ID 59
5.1 Requirements 59
5.2 Background 60
5.3 Test Procedures 60
5.4 Requirements Compliance 60

6 Fast-Track Certification 61
6.1 Requirements 61
6.2 Fast Track Permission 64
6.3 Test Procedures 64
6.4 Requirements Compliance 64

7 Additional / Optional Data 65

7.1 Inked Card Comparison for New Designs 65
7.2 Stress Imagery 66
7.2.1 Simulated Sunlight Collection Guidance 66
7.2.2 Black X Collection Guidance 67
7.3 Metadata Documentation 67

vii
List of References 69

Appendix A - Livescan Testing Notes 71

Appendix B - Commercial Sources for Targets 73

Appendix C - Geometric Accuracy Measurement 77

Appendix D - Construction of Stratified Test Card Set 87

Glossary 91

viii
 LIST OF FIGURES

FIGURE PAGE

1-1 Standard Ten-Print Fingerprint Card (FD-258) 4

2-1 Linearity Requirement for Scanner Input/Output 9

2-2 Layout of Four 4 x 4 Inch Ronchi Targets 12

2-3 Bar Center to Bar Center Measurements 13

2-4 Along-Bar Geometric Accuracy Assessment 14

2-5 Specification Scanner MTFs (Sine Wave Target) and CTFs (Bar Target) 17

2-6 Example Layout for 15-Bar Target in Identification Flats Live Scanner 21

2-7 Example Layout for 2 x 3 Inch Sine Wave Target (M13-60-1X)

in Card Scanner 22

2-8 Example Layout for Bi-Directional Bar Target Centered in Roll Capture Area 23

2-9 Composite of Two Uniform Gray Targets 30

2-10 Example Measurements for Adjacent Row Uniformity Assessment 32

2-11 Example Measurement Boxes for Gray Range Assessment 36

4-1 Printer Test Target (TGT) 44

4-2 Print Dynamic Range Measurement 47

4-3 Correct Orientation of Print Ronchi Bars with Respect to Scanning Array 49

4-4 Auxiliary Information in Ten-Print Card Format Print 54

7-1 Suggested Set-up to Simulate Full Sunlight Exposure 67

C-1 Illustration of Definitions Used For Ronchi Target Computations 78

C-2 Example of Calculating the Edge Lines of a Bar Segment 80

ix
C-3 The Distance Between a Point and a Line 81

C-4 The Distance Between Two Bar Centers is the “One Bar Distance
Measurement” 81

C-5 Examples of First and Last Full-Width Bar Segments in a

Measurement Strip 82

C-6 Example of Along-Bar Distortion Measurement 85

x
 LIST OF TABLES

TABLE PAGE

1-1 Preferred Capture Sizes 5

2-1 Geometric Accuracy (inches) and Resolution (ppi) Requirements 15

2-2 MTF Requirement Using Sine Wave Target 16

2-3 CTF Requirement Using Bar Target 16

2-4 Minimum Number of Target Bars 18

2-5 Minimum Target Layouts 20

2-6 Example 50 Card Test Results 37

4-1 Geometric Accuracy Requirements 48

4-2 Geometric Accuracy Tests 50

5-1 Mobile ID IQS Requirements 59

6-1 Fast Track Certification Procedures (Common Scenarios) 63

B-1 Test Targets 73

D-1 Strata, Gray Scale Ranges, Sample Sizes for FCMF 100 Card Test Set 87

D-2 Gray Scale Ranges in 100 Card FCMF Test Set 88

D-3 Gray Scale Ranges for Card Test Set 89

xi
(blank page)

xii
 SECTION 1

INTRODUCTION

This document presents the test procedures that are used to verify the FBI's Integrated
Automated Fingerprint Identification System’s (IAFIS) Image Quality Specifications
(IQS), as defined in Appendix F of [EBTS]. The IQS applies to: (1) systems which scan
and capture fingerprints3 in digital, softcopy form, including hardcopy scanners such as
ten-print card scanners and latent print scanners, and live scan devices, altogether called
“fingerprint scanners”; and (2) systems utilizing a printer to print digital fingerprint
images onto paper or card stock, called “fingerprint printers”.

The terms "paper scanner" and "live scanner" are used in this document to refer to the
two basic types of fingerprint scan/capture devices. The hardcopy input to a paper
scanner is a fingerprint on a reflective paper substrate, such as an inked fingerprint on
card stock, or a latent print on photographic paper. In a live scanner, fingerprints are
directly captured from a subject's fingers and output in digital form. When the term
"scanner" is used in this document with no qualifier, it refers to both paper scanners and
live scanners.

Verification of compliance of a scanner or printer to the full set of IQS requirements is

primarily performed by the test method; i.e., verification through systematic exercising of
the item with sufficient instrumentation to show compliance with the specified
quantitative criteria. A few requirements are verified by the inspection method; i.e.,
verification of requirements by visual examination of the item or review of descriptive
documentation.

The requirements for Fast Track certification testing are a subset of the full testing
described in this document; they are described in section 6. Data requests that either fall
outside the current requirements or only apply in very specific instances are documented
in section 7.

General Test Attributes

• All required testing is the responsibility of the vendor who is seeking certification.

• All testing should be performed on a single, representative unit of the product/model for
which certification is being sought. The test unit should be operated in its normal

3The term “fingerprint” in this document may also include palmprint, whole hand print, or a print from
another part of the human body.

1
operating mode, to the degree consistent with obtaining the test images4. For example, a
scanner should not be tested at slower than normal operating speeds to meet geometric
accuracy requirements.

• For scanner testing, the digital test images are to be supplied in an uncompressed format
such as raw, TIFF, or PGM formats, at 8 bits per pixel (8 bpp or 256 gray-levels). Do not
submit compressed images and do not supply uncompressed images that were previously
compressed via a lossy compression; e.g., do not supply images that are decompressed
WSQ images.

• The vendor is encouraged to analyze its test data results before submitting the test data
to the FBI; test data analysis software (freeware) is available through the FBI for this
purpose (IQS Test Tools CD). The vendor has the option of submitting its test results
and any additional information, such as discussion of results or technical/design
information on the device that is relevant to the testing. The vendor also has the option to
utilize its own test analysis software, or modify the FBI-supplied software. However, the
vendor should expect that an independent analysis of the test data will be performed by
the FBI or its supporting organizations. Therefore, the test images (for scanner) or test
prints (for printer) must be submitted to the FBI. Analysis by the FBI or its supporting
organizations will primarily rely on the same IQS Test Tools that are made available to
the vendor.

• If the test data analysis by the FBI or its supporting organizations indicates
noncompliance with any of the IQS requirements, the vendor will be informed of the
specific deficiencies and given an opportunity to correct the deficiencies and submit new
test data.

• If the vendor has good reason to believe that certain test procedures given in this
document are not applicable to the vendor’s device, or cannot be applied without
modifying the device, or if other test procedures the vendor has confidence in can be
shown to be equivalent to the test procedures given in this document and are better suited
to testing the particular device, then the vendor may initiate communications with the FBI
in this regard, which may result in acceptable modified, device-tailored test procedures.
Potential test modifications are particularly relevant to livescan devices; refer to
Appendix A - Livescan Testing Notes.

4 However, the "adaptive processing" that may be applied to fingerprint capture (IQS section 2.6) should
not be applied when scanning test targets. If, due to device design features, it is necessary to apply
fingerprint adaptive processing to the test target scans, the adaptivity (settings, level) must be invariant over
the entire scanned area of the test target.

2
Test Targets

Following is a listing of the commonly used scanner test targets; refer to Appendix B -
Commercial Sources for Targets, for obtaining these targets.

• Multiple parallel bar target at 1.0 cycles per millimeter (cy/mm) for geometric accuracy
and pixels per inch resolution tests.

• Sine wave target with gray patches for Modulation Transfer Function (MTF) and
linearity tests.

• Uniform dark gray and light gray targets for signal-to-noise ratio and gray-level
uniformity.

Scanner testing also includes scanning a set of fingerprints. For paper scanners, a
specific test set of inked fingerprint cards is supplied by the FBI. For live scanners, the
vendor is responsible for supplying a set of livescans.

Fingerprint printers are tested with a digital test target and a set of digitized test
fingerprints; both of these are on the IQS Test Tools CD.

Image Capture Areas

At a minimum, a ten-print card scanner shall be capable of capturing an area of at least

5.0 x 8.0 inches, which captures all 14 printblocks, either each printblock as a separate
image, or all printblocks together as a single image. Figure 1-1 shows the standard ten-
print applicant fingerprint card, FD-258, with dimensions of major components5. When a
specific test procedure in this document refers to a "5 x 8 inch scan area", it is referring to
the 14 print block area, as illustrated in Figure 1-1.

5 The older FD-249 card has the same overall size and printblock dimensions as FD-258, but a different
format/layout for the text blocks.

3
 8″!

LEAVE BLANK! TYPE OR PRINT ALL INFORMATION IN BLACK! FBI! LEAVE BLANK!
LAST NAME NAM! FIRST NAME MIDDLE NAME !
APPLICANT!

SIGNATURE OF PERSON FINGERPRINTED! ALIASES AKA! O

R!
I!
RESIDENCE OF PERSON FINGERPRINTED! DATE OF BIRTH DOB!
Month Day Year!

3″!
CITIZENSHIP CTZ! SEX! RACE! HGT.! WGT.! EYES! HAIR! PLACE OF BIRTH POB!
DATE! SIGNATURE OF OFFICIAL TAKING FINGERPRINTS!

YOUR NO. OCA!

LEAVE BLANK!
EMPLOYER AND ADDRESS!
FBI NO. FBI!

ARMED FORCES NO. MNU! CLASS_______________________________________________!

!
REASON FINGERPRINTED!
SOCIAL SECURITY NO. SOC!
!
REF._________________________________________________!
MISCELLANEOUS NO. MNU!

1.6″! 1.5″!

1. R THUMB ! 2. R. INDEX! 3. R. MIDDLE! 4. R. RING! 5. R. LITTLE!

5″!
6. L THUMB ! 7. L. INDEX! 8. L. MIDDLE! 9. L. RING! 10. L. LITTLE!

0.8″!
3.2″!
2.0″!
LEFT FOUR FINGERS TAKEN SIMULTANEOUSLY! L. THUMB! R. THUMB! RIGHT FOUR FINGERS TAKEN SIMULTANEOUSLY!

Figure 1-1. Standard Ten-Print Fingerprint Card (FD-258)

4
In terms of individual printblocks, Table 1-1 gives the preferred capture sizes, applicable
to both card scan and live scan systems [EBTS, ANSI/NIST], with the exception that,
when scanning fingerprint cards, the card form dimensions take precedence.

Table 1-1. Preferred Capture Sizes

Preferred Width Preferred Height

(inches) (inches)
roll finger 1.6* 1.5
plain thumb 1.0 2.0
plain 4-fingers 3.2 2.0
(sequence check)
plain 4-fingers 3.2 3.0
(identification flat)
full palm 5.5 8.0
half palm** 5.5 5.5
writer's palm 1.75 5.0

* A live scanner must be capable of capturing at least 80% of full roll arc length,
where full roll arc length is defined as arc length from nail edge-to-nail edge.
** Although larger sizes are preferred, minimum acceptable half-palm dimensions
are 5.0 x 5.0 inches.

Image Gray-Level Quantization

The final output of all test target scans and all test (and operational) fingerprint scans,
shall be gray-level quantized to 8 bpp (256 gray-levels).

Point-of-Contact

Questions or concerns regarding the IQS requirements, test targets, test procedures,
acceptability of alternate targets/procedures, certification procedures, or availability of
test analysis software (Test Tools CD), can be addressed to mtf@mitre.org, which acts on
the FBI’s behalf for purposes of IQS testing.

Questions and concerns may also be addressed to Eric Phillips of the FBI:

Eric M. Phillips

Telephone: 304-625-4531
Email: eric.phillips@ic.fbi.gov

5
(blank page)

6
 SECTION 2

FINGERPRINT SCANNER

2.1 LINEARITY

2.1.1 Requirements

When measuring a stepped series of uniform target reflectance patches (“step tablet”) that
substantially cover the scanner’s gray range, the average value of each patch shall be
within 7.65 gray-levels of a linear, least squares regression line fitted between target
reflectance patch values (independent variable) and scanner output gray-levels
(dependent variable).

2.1.2 Background

All targets used in IQS compliance verification are expected to be scanned with the
scanner operating in a linear input/output mode. Linearity enables valid comparisons of
test measurements with requirements; e.g., a system’s spatial frequency response in terms
of Modulation Transfer Function is, strictly speaking, a linear systems concept. Linearity
also facilitates comparisons between different scanners through the “common ground”
concept. For fingerprint scans, linearity produces a pristine image in a common reference
base. From this base, users such as an Automated Fingerprint Identification System
(AFIS) or fingerprint examiners working in softcopy, can then apply linear/non-linear
processing, as needed for specific purposes, with the benefit that they are always able to
get back to the base image.

However, in atypical cases, linearity may be waived for test target scans; i.e., a small
amount of smooth, monotonic nonlinearity may be acceptable when it is substantially
impractical and unrepresentative of operational use to force linearity on the scanner under
test. Such cases require the submission of documentation along with the waiver request.

It is recognized that the ten-print card, latent photo, or live finger input to the scanner
may have less than ideal characteristics, in terms of average reflectance, discontinuities in
average reflectance, low contrast, and/or background clutter. Such problems may
sometimes be minimized by applying nonlinear gray-level processing to the scanner-
captured image. For example, reduction of white background clutter surrounding a
fingerprint leads to more of the available image compression bits being allocated to the
fingerprint itself, which results in higher quality when the fingerprint image is
decompressed. For these reasons, linearity is not a requirement for the operational or test
fingerprint scans.

7
2.1.3 Target

For a paper scanner, or any live scanner that can image a continuous tone target, a step
tablet covering the dynamic range of the scanner shall be used as the target; e.g., the step
tablet component of the sine wave target used for MTF assessment fulfills this purpose
(see Figure 2-7).

For live scanners that cannot image a continuous tone target, the appropriate target is
dependent on the live scanner’s design and imaging capabilities. For example, it may be
necessary to produce a series of images of neutral density filters inserted into the optical
path, or a series of images of a uniform (blank) platen with variation of exposure time at
constant intensity. Whatever target/method is used, it must produce at least 9 unique
gray-level steps covering the dynamic range of the scanner.

See Appendix B for commercial sources for step tablets.

2.1.4 Test Procedures

A linear, least squares regression is run between the step-averaged target reflectance or
transmission values (input predictor variable), and the corresponding step-averaged
scanner output gray-levels (output response variable), producing the equation of a
straight line:

Predicted OutputGrayLevel = Slope x TargetReflectance + Intercept

The deviation of each step-averaged scanner output step gray-level from the linear, least
squares regression line of best fit is noted, as illustrated in Figure 2-1.

The linearity requirement can be verified with the sinemtf software, which is on the IQS
Test Tools CD. The latest version of sinemtf is always posted to the internet at:

http://www.mitre.org/tech/mtf

2.1.5 Requirements Compliance

For each of the target steps, the absolute value of the difference between the step-
averaged scanner output gray-level and the linear regression predicted gray-level, shall be
no greater than 7.65 gray-levels.

8
 250
Image
Gray
200

150
<
_ 7.65

100

50

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Targe t Reflectance

Figure 2-1. Linearity Requirement for Scanner Input/Output

2.2 RESOLUTION AND GEOMETRIC ACCURACY

2.2.1 Requirements

Resolution:
The scanner’s final output fingerprint image shall have a resolution, in both sensor
detector row and column directions, in the range: (R – 0.01R) to (R + 0.01R). The
magnitude of “R” is either 500 ppi or 1000 ppi; a scanner may be certified at either one,
or both, of these resolution levels. The scanner’s true optical resolution shall be greater
than or equal to R.

Across-Bar Geometric Accuracy:

When scanning a 1.0 cy/mm, multiple parallel bar target, in both vertical bar and
horizontal bar orientations, the absolute value of the difference (D), between the actual
distance across parallel target bars (X), and the corresponding distance measured in the
image (Y), shall not exceed the following values, for at least 99% of the tested cases in
each printblock measurement area and in each of the two directions.

9
 for 500 ppi scanner:
D ≤ 0.0007, for 0.00 < X ≤ 0.07

D ≤ 0.01X, for 0.07 ≤ X ≤ 1.50

for 1000 ppi scanner:

D ≤ 0.0005, for 0.00 < X ≤ 0.07

D≤ 0.0071X, for 0.07 ≤ X ≤ 1.50

where:
D = |Y-X|
X = actual target distance
Y = measured image distance
D, X, Y are in inches

Along-Bar Geometric Accuracy:

When scanning a 1.0 cy/mm, multiple parallel bar target, in both vertical bar and
horizontal bar orientations, the maximum difference in the horizontal or vertical
direction, respectively, between the locations of any two points within a 1.5 inch segment
of a given bar image, shall be less than 0.016 inches for at least 99% of the tested cases in
each printblock measurement area and in each of the two orthogonal directions.

2.2.2 Background

A multiple parallel bar target refers to a Ronchi target, which consists of an equal-width
bar and space square wave pattern with high contrast ratio and sharp edge definition.

For a 500 ppi system, the resolution must be between 495.0 and 505.0 ppi; for a 1000 ppi
system, the resolution must be between 990.0 and 1010.0 ppi. The scanner’s true optical
resolution may be greater than the required resolution, in which case rescaling down to
the required resolution is performed for final output. However, the scanner’s true optical
resolution cannot be less than the required resolution; i.e. “upscaling”, from less than the
required ppi resolution, to the required ppi resolution, is not allowed.

Across-bar geometric accuracy is measured across imaged 1.0 cy/mm Ronchi target bars
that substantially cover the total image capture area. The 500 ppi requirement
corresponds to a positional accuracy of ± 1.0% for distances between 0.07 and 1.5
inches, and a constant ± 0.0007 inches (1/3 pixel) for distances less than or equal to 0.07
inches. The 1000 ppi requirement corresponds to a positional accuracy of ± 0.71% for
distances between 0.07 and 1.5 inches, and a constant ± 0.0005 inches (1/2 pixel) for
distances less than or equal to 0.07 inches.

10
Along-bar geometric accuracy is measured along the length of imaged, 1.0 cy/mm
Ronchi target bars that substantially cover the total image capture area. For a given
horizontal bar, for example, the maximum difference between bar center locations (in
vertical direction), determined from bar locations measured at multiple points along a 1.5
inch bar segment length, is compared to the maximum allowable difference requirement
(analogously for vertical bar). This requirement is to ensure that pincushion or barrel
distortion over the primary area of interest; i.e., a single fingerprint, is not too large.

2.2.3 Target

Scanner resolution and geometric accuracy are measured using a precision Ronchi target
having a constant spatial frequency of 1.0 cy/mm; i.e., the combined width of one black
bar plus one adjacent white space is 1.0 mm, which is one cycle, or one period. This
target is available on a reflective white mylar substrate, which is suitable for use in
testing paper scanners. Live scanners may require use of a different substrate, such as a
chrome-on-glass Ronchi target. Ronchi targets on mylar and glass substrates are
commercially available from several vendors – see Appendix B.

For a card scanner, testing shall be performed over a 4 x 8 inch area, in both vertical and
horizontal directions. The required two positions and two orientations of a typically
available 4 x 4 inch Ronchi target are illustrated in Figure 2-2. The preferred method is
to mount two 4 x 4 inch Ronchi targets on a single ten-print card stock, then, only two
scans are required, with the card rotated 180 degrees for the second scan. Alternatively, a
single 4 x 4 inch Ronchi target can be alternately placed in the two locations and two
orientations, but this requires four separate scans and does not capture both directions in
a single scan.

The Scanner Image Quality Test (SIQT) target was designed for testing fingerprint card
scanners. If this target is included in the FBI supplied ten-print test card set (see section
2.6), or is otherwise available, then it should also be scanned on the card scanner. This
target contains horizontal and vertical Ronchi bars and serves as a good adjunct
measurement of geometric accuracy, particularly useful when marginal results are
obtained with the required 4 x 4 inch Ronchi targets. See Appendix B for commercial
sources for the SIQT target.

For a live scanner, resolution and geometric accuracy testing shall be performed over at
least 70% of the scanner’s capture area, in both vertical and horizontal directions.

11
Figure 2-2. Layout of Four 4 x 4 Inch Ronchi Targets
(A & D - Horizontal Bars, B & C - Vertical Bars,
nominal number of 10 measurement strips per target)

12
2.2.4 Test Procedures

2.2.4.1 Resolution and Across-Bar Geometric Accuracy

The goal is to acquire measurements that comprehensively cover a Ronchi target image,
in continuous, 1/4 inch wide measurement strips running the height (vertical direction
measurements) or width (horizontal direction measurements) of the target image. Target
blemishes or dust may negate successful measurement in isolated areas; in such cases it
may be necessary to segment a long, single measurement strip depicted in Figure 2-2 into
a number of shorter strips, or shift measurement locations.

For each measurement strip, measurements are taken across all independent one bar
distances and six bar distances. First, the scanner output resolution is measured over 6
Ronchi bar cycles (“6-bar distance”); this verifies compliance with the pixels per inch
requirement and establishes a pixels per inch value that can be used to convert
subsequent geometric accuracy measurements to inches. Next, measurements are made
to test the geometric accuracy over a short distance of one Ronchi bar cycle (“1-bar
distance”). Finally, measurements are made to test the geometric accuracy over a longer
distance of 6 Ronchi bar cycles. All distances for geometric accuracy are measured from
bar center to bar center in pixel units, as illustrated in Figure 2-3. The pixel
measurements are then converted to inches using the previously computed average
resolution for the given measurement strip. A bar center is located by first detecting that
bar's left and right edges along a 1/4 inch edge height (for vertical bar in scanner
assessment), and then bisecting the bar’s two edges, taking skew angle into account.

1 bar cycle

0.25
inches

6 bar cycles

Figure 2-3. Bar Center to Bar Center Measurements

2.2.4.2 Along-Bar Geometric Accuracy

This test for distortion utilizes the same Ronchi targets used for across-bar geometric
accuracy assessment. All distances for along-bar geometric accuracy are first measured
from local bar center location to local bar center location in pixel units. The difference in

13
fractional rows between two bar centers is converted to inches by dividing by the average
ppi of the two corresponding measurement strips. Figure 2-4 illustrates a single
measurement for a single horizontal bar. In this example, the maximum vertical
deviation “H” occurs for two bar center locations that are spaced at the maximum
measurement distance of 1.5 inches (in horizontal direction), but note that the maximum
vertical deviation could occur at bar center locations that are less than 1.5 inches apart.

1.5 inches
H

Figure 2-4. Along-Bar Geometric Accuracy Assessment

(Vertical Height “H” is Test Sample Value)

The geometric accuracy and resolution requirements can be verified with the geo
software, supported by creategeofile and viewgeo software, which are on the IQS Test
Tools CD. See Appendix C for a detailed description of the computations performed in
geo.

2.2.5 Requirements Compliance

The resolution requirement is complied with if the average resolution, in each printblock
measurement area, and in each direction, is within 1.0 percent of the required scanner
resolution, as delineated in Table 2-1.

The across-bar geometric accuracy requirement is complied with if at least 99.0 percent
of the tested cases, in each printblock measurement area, and in each direction, are within
the minimum and maximum distance limits defined in Table 2-1.

The along-bar geometric accuracy requirement is complied with if at least 99.0 percent of
the test measurement values (“H” in Figure 2-4), in each printblock measurement area
and in each direction, are less than 0.016 inches.

14
 Table 2-1. Geometric Accuracy (inches) and Resolution (ppi) Requirements

Measurement Correct Directional Requirement Met if:

Value
1-bar distance @ 500 ppi 0.03937 ≥ 99% in range: 0.03867 to 0.04007

1-bar distance @ 1000 ppi 0.03937 ≥ 99% in range: 0.03887 to 0.03987

6-bar distance @ 500 ppi 0.23622 ≥ 99% in range: 0.23386 to 0.23858

6-bar distance @ 1000 ppi 0.23622 ≥ 99% in range: 0.23454 to 0.23790

500 ppi resolution 500.0 avg in range: 495.0 to 505.0

1000 ppi resolution 1000.0 avg in range: 990.0 to 1010.0

2.3 SPATIAL FREQUENCY RESPONSE

2.3.1 Requirements

The spatial frequency response shall be measured using a continuous tone sine wave
target, denoted as Modulation Transfer Function (MTF) measurement, unless the scanner
cannot obtain adequate tonal response from this target, in which case a bi-tonal bar target
shall be used to measure the spatial frequency response, denoted as Contrast Transfer
Function (CTF) measurement. When measuring the sine wave MTF, it shall meet or
exceed the minimum modulation values given in Table 2-2, in both the detector row and
detector column directions, and over any region of the scanner's field of view. When
measuring the bar CTF, it shall meet or exceed the minimum modulation values defined
by equation 2-1 or equation 2-2 (whichever applies), in both the detector row and
detector column directions, and over any region of the scanner's field of view. CTF
values computed from equations 2-1 and 2-2 for nominal test frequencies are given in
Table 2-3. The specification MTFs and CTFs are plotted in Figure 2-5.

None of the MTF or CTF modulation values measured at specification spatial frequencies
shall exceed 1.05.

The output sine wave image or bar target image shall not exhibit any significant amount
of aliasing.

15
 Table 2-2. MTF Requirement Using Sine Wave Target

Frequency Minimum Modulation for Minimum Modulation Maximum Modulation

(cy/mm) 500 ppi Scanner for 1000 ppi Scanner
1.0 0.905 0.925
2.0 0.797 0.856
3.0 0.694 0.791
4.0 0.598 0.732
5.0 0.513 0.677
6.0 0.437 0.626
7.0 0.371 0.579
1.05
8.0 0.312 0.536
at all frequencies
9.0 0.255 0.495
10.0 0.200 0.458
12.0 0.392
14.0 0.336
16.0 0.287
18.0 0.246
20.0 0.210

Note: Testing at 7 and 9 cy/mm is not a requirement if these frequency patterns are absent from
the sine wave target.

Table 2-3. CTF Requirement Using Bar Target (Nominal Test Frequencies)

Frequency Minimum Modulation Minimum Modulation Maximum Modulation

(cy/mm) for 500 ppi Scanner for 1000 ppi Scanner
1.0 0.948 0.957
2.0 0.869 0.904
3.0 0.791 0.854
4.0 0.713 0.805
5.0 0.636 0.760
6.0 0.559 0.716
7.0 0.483 0.675
1.05
8.0 0.408 0.636
at all frequencies
9.0 0.333 0.598
10.0 0.259 0.563
12.0 0.497
14.0 0.437
16.0 0.382
18.0 0.332
20.0 0.284

Note: Testing at or near 7 and 9 cy/mm is a requirement when using a bar target.

16
 1
Modulation
0.9

0.8

0.7

0.6

0.5
CTF1000

0.4

0.3 CTF500

0.2 MT F1000
MT F500

0.1

0
0 2 4 6 8 10 12 14 16 18 20
Spatial Frequency (cy/mm)

Figure 2-5. Specification Scanner MTFs (Sine Wave Target) and CTFs (Bar Target)

It is not required that the bar target contain the exact frequencies listed in Table 2-3;
however, the target does need to cover the listed frequency range and contain bar patterns
close to each of the listed frequencies. The following equations are used to obtain the
minimum acceptable CTF modulation values when using bar targets that contain
frequencies not listed in Table 2-3.

500 ppi scanner, for f = 1.0 to 10.0 cy/mm:

CTF = 3.04105E-04*f 2 – 7.99095E-02*f + 1.02774 (2-1)

1000 ppi scanner, for f = 1.0 to 20.0 cy/mm:

CTF = -1.85487E-05*f 3 + 1.41666E-03*f 2 – 5.73701E-02*f + 1.01341 (2-2)

For a given bar target, the specification frequencies include all of the bar frequencies
which that target has in the range 1 to 10 cy/mm (500 ppi scanner) or 1 to 20 cy/mm
(1000 ppi scanner).

17
2.3.2 Target

For a paper scanner, a commercially available, reflective sine wave target is used (this
target can also be used to verify the linearity requirement in section 2.1).

For a live scanner, it may be possible to use a reflective or transmissive sine wave target,
and this is the first choice. If the live scanner is not compatible with capturing a sine
wave target and producing a continuous gray tone image, then a bi-tonal, black & white,
bar target is used. The bar target must have the following properties:

• Bars must cover the frequency range from 1 cy/mm up to the scanner output Nyquist
frequency of 10 cy/mm for a 500 ppi scanner, or 20 cy/mm for a 1000 ppi scanner.
The bar target frequency patterns must be within 0.49 cy/mm of each of the
frequency increments: 1, 2, 3, 4, 5, 6, 7, 8, 9 cy/mm, and within 0.25 cy/mm of 10
cy/mm.

• The bar target must contain at least one very low frequency component; i.e., a large
square, single bar, or series of bars whose effective frequency is no greater than 3%
of the scanner output Nyquist frequency. This low frequency component is used in
normalizing the CTF; it must have the same density on the target as the other target
bars. For example, a scanner with 500 ppi output resolution would require that the
target contain at least one bar whose width is at least 1.7 mm.

• In the frequency range of 1 cy/mm up to the scanner output Nyquist frequency, there
must be an adequate number of target bars at each frequency; a minimum number is
defined in Table 2-4. A minimum number of bars is needed in order to ensure
capturing the optimum phase between scanner sensor array and target bars, in order
to have enough samples available for accurate measurement of aliasing, and in order
to obtain an accurate measure of modulation.

Table 2-4. Minimum Number of Target Bars

Target Frequency Minimum Number of Parallel Bars

≤ 3% of Nyquist 1
> 3% of Nyquist to 1 cy/mm 4
> 1 to 4 cy/mm 5
> 4 cy/mm 10

Notes: Bar length must be at least 5 times bar width.

Width of space between parallel bars equals bar width.

18
 • The bar target may be either a commercially purchased standard bar target design
(see Appendix B), or may be specially designed/fabricated for the IQS testing. The
bar target may have any suitable substrate (paper, mylar, film, glass), depending on
the imaging requirements of the specific live scanner. Bar targets seldom include a
step tablet; the linearity requirement must therefore be tested with other test targets
and other methods, as appropriate for the particular live scanner design.

2.3.3 Test Procedures

1) The acceptable number of locations and orientations of a target depends principally on

target size, relative to the scanner’s total capture area, and whether or not the target
contains frequency patterns in one direction, or in two orthogonal directions. In every
case, however, the target must be oriented such that the scanner’s spatial frequency
response can be measured in both horizontal and vertical directions (scanner detector
array row and column directions). The following subsections 2.3.3(a), 2.3.3(b) and
2.3.3(c) present guidelines for acceptable layouts for some common target designs, as
summarized in Table 2-5. Other target designs or potentially unique aspects of image
capture areas may require other layouts.

a) If the area of the target is less than 25% of the scanner's total capture area, then an
acceptable layout consists of one vertical and one horizontal target, each located
off-center. Example layouts are illustrated in Figure 2-6 for bar targets in an
Identification Flats live scanner, and in Figure 2-7 for 2 x 3 inch sine wave targets
in a ten-print card scanner.

b) If the area of the target is greater than or equal to 25% of the scanner's total
capture area, then an acceptable layout consists of one target centered in the
image capture area, capturing both vertical and horizontal bar orientations (via
separate scans, if necessary). Figure 2-8 is an example layout of a single bar
target with both vertical and horizontal components in a roll capture area.

c) If the scanner has a large total capture area (such as half-palm or Identification
Flats) with a small subsection devoted to roll captures, then the target must be
imaged in the critical roll capture area (via separate scans, if necessary) as well as
across the full capture area as required in 2.3.3(a) or 2.3.3(b).

19
 Table 2-5. Minimum Target Layouts

Target Area as % Target Contains Vertical (V) Target Contains Vertical (V)
of Total Capture and Horizontal (H) or Horizontal (H)
Area Components Components
Single scan of two off-center
< 25% Single scan of two off-center targets, one V target & one H
targets; or separate scan of each target; or separate scans of one
of two off-center targets off-center V target and one off-
center H target
Separate scans of one centered
≥ 25% Single scan of centered target V target and one centered H
target

Notes:
- The target area is the area of the target itself; e.g., area covered by sine patterns and
surrounding gray patches, or the area of an array of bars, regardless of the target substrate
dimensions.
- “Centered target” means that the center of the target coincides with the point at which the
scanner’s optical axis intersects the scanner’s object plane, which usually coincides with the
center of the capture area.
- When two off-center targets are used, they are expected to be on opposite sides of the center
of the capture area.

20
 3.2"

3.0"

1.0"

Figure 2-6. Example Layout for 15-Bar Target in Identification Flats Live Scanner
(Single Target Area is 10% of Capture Area)

21
 8ʺ

LEAVE BLANK TYPE OR PRIN T A LL INFORMA TION IN FBI LEAVE BLANK

BLACK FIRST NAME MIDD LE NAME
LAST NA ME
APPLICAN NAM

T
SIGNATUR E OF PERSON ALIASES O
FIN GER PRIN TED AKA R
I
RESID ENCE OF PERSON DATE OF B IRTH D OB
FIN GER PRIN TED Month D ay
Year

3ʺ
CITIZENSHIP SE RAC HGT WGT. EYE HAI PLACE OF BIR TH POB
CTZ X E . S R
DAT SIGNATUR E OF OFFIC IA L TAKIN G
E FIN GER PRIN TS
YOUR NO.
EMPLOYER AND
OCA LEAVE BLANK
ADDRESS
FBI N O.
FBI

ARMED FORCES N O. CLASS__________________________________________

MNU
REASON _
FIN GER PRIN TED SOC IAL SECUR ITY N O.
SOC
REF.____________________________________________
MISC ELLANEOUS N O.
MNU
_

2. R.
INDEX

4. R. 5. R.
1. R RING LITTLE
3. R. MIDDLE
THUMB
1.9ʺ

5ʺ
7. L. IND EX

6. L 8. L. 9. L.
2.8ʺ 10. L.
THUMB MIDD LE RING LITTLE

LEFT FOUR FINGERS TAK EN L. R. RIGH T FOUR FIN GER S TAKEN

SIMU LTANEOUSLY THUMB THUMB SIMU LTANEOUSLY

Figure 2-7. Example Layout for 2 x 3 Inch Sine Wave Target (M13-60-1X)
in Card Scanner (Single Target Area is 8% of Capture Area)

22
 1.6″

0.8″
1,3 4,2

5,7 8,6
9 10
1.5″

1,3 2,4 9,10 5,7 6,8

Figure 2-8. Example Layout for Bi-Directional Bar Target Centered

in Roll Capture Area (Target Area is 27% of Roll Capture Area)

23
The following procedures, sections 2.3.3(2) to 2.3.3(7), correspond to input requirements
and computations performed by the sinemtf software for MTF and CTF assessment. This
software, together with detailed documentation, can be found on the IQS Test Tools CD;
the most recent version of sinemtf can also be downloaded from the internet at:

http://www.mitre.org/tech/mtf

2) A target data file is prepared for each target model, which includes the relative
locations within the target of each frequency pattern, the modulation value of each
frequency pattern, and gray patch densities/locations (if target contains gray patches).

For a sine wave target, the target manufacturer supplies the target modulations and gray
patch densities for the specific target serial number purchased. The supplied
"compensated modulation" values are used for target modulation, because these values
have been corrected for the instrument (microdensitometer) used by the manufacturer to
calibrate the target. [The supplied “peak-to-peak” modulation is not used, because these
values have not been corrected.] The constructed target data file is submitted along with
the target scans. As verification that “compensated modulation” has been used, at least a
portion of the target manufacturer’s data sheet for the specific serial-numbered sine wave
target should also be submitted; e.g., the portion which includes target modulation data at
6 cy/mm.

3) The target is positioned on the scanner platen, aligned to within 0.5 degrees of the
horizontal and vertical (scanner detector array rows and columns), and scan capture
parameters are appropriately set. Note that due to the medium contrast of sine wave
targets it is relatively easy to avoid saturation upon imaging; i.e., easy to avoid the sine
wave peaks in the image being pinned at gray = 255 or the sine wave valleys being
pinned at gray = 0. On the other hand, a bar target is very high contrast and is more
prone to producing a saturated image. Adjustments in scanner parameters (illumination,
gain settings, etc.) should be made, as necessary to avoid saturation. If some bar target
saturation is unavoidable, it is better to saturate the white spaces at gray = 255 rather than
the black bars at gray = 0.

4) The target is scanned. For any case in which more than one target is captured in a
single scan, the single scan should be submitted, not cropped or segmented into separate
target images.

5) The digital image of the scanned target is displayed and the upper left, upper right, and
lower left corners are located (row and column in pixel units). Sinemtf is then run, with
inputs consisting of the 3 image reference corners, target data file, digital image, and
various user-selectable options displayed at runtime. Sinemtf first computes the pixels
per inch and alignment angle (“skew angle”) in both vertical and horizontal directions for
the given target image. This data, together with the relative locations of each target

24
pattern (from input data file), is then used to establish the row and column location of
each density patch and each sine wave (or bar) frequency pattern within the image.

6) The single, representative sine wave (or bar cycle) modulation in each imaged
frequency pattern is determined from the sample modulation values collected from within
that pattern. The sample modulation values are computed from the maximum and
minimum levels corresponding to the peak and adjacent valley in each sine wave period
(or bar cycle). These maximum and minimum levels represent the image gray-levels that
have been locally averaged in a direction perpendicular to the sinusoidal (or bar cycle)
variation; for sine waves, these maximum and minimum gray-levels are then mapped
through a calibration curve into target reflectance space where the image modulation is
computed as,

peak_image_modulation = (maximum - minimum) / (maximum + minimum)

The calibration curve is the curve of best fit between the image gray-levels of the density
patches in the sine wave target and the corresponding target reflectance values. The
scanner MTF at each frequency is then defined as:

MTF = peak_image_modulation / target_modulation

For CTF assessment using a bar target, the modulations are determined in image space,
normalized by the image modulation at zero frequency6, instead of using a calibration
curve. The scanner CTF at each frequency is then defined as:

CTF = peak_image_modulation / zero_frequency_image_modulation

[Unless specifically measured, a bar target’s modulation values are assumed to be

constant, at all frequencies up to the scanner output Nyquist frequency.]

A maximum modulation of 1.05 is defined to discourage the use of sharpening, edge

enhancement, or MTF-boost filters in the signal processing that produces the image;
these filters tend to make the MTF go above 1.0 at some frequencies. These filters are
unwanted for two reasons: (1) they have the potential for introducing artifacts that are not
in the original fingerprint; i.e., a pristine digitized fingerprint is desired, and (2) they can
introduce high frequency energy which prevents the creation of the smooth edge surfaces
that are compatible with wavelet compression techniques (WSQ, JPEG-2000).

6In this context, “zero frequency” refers to any single or multiple bar pattern whose spatial frequency is no
greater than 3% of the scanner output Nyquist frequency.

25
7) Aliasing Measurement - Aliasing is measured because it is a potential source of
unwanted image artifacts; e.g., pronounced aliasing may produce false detail in the
image, such as a pseudo-ridge pattern. Aliasing is measured in sinemtf by computing a
sequence of one-dimensional discrete Fourier transforms (DFT) of the row-averaged
gray-levels in each frequency pattern. If the relative strengths of side lobes (harmonics)
compared to the main lobe are too large, then aliasing due to nonuniform decimation is
called out. If the location of the main lobe is not at the correct frequency, for a given
pattern with known fundamental frequency, then aliasing due to upscaling is called out.
Aliasing due to decimation is usually not entirely avoidable when the scanner’s true
optical resolution is greater than the required 500 ppi or 1000 ppi, because of the
rescaling algorithm (resampling and interpolation) that must be applied in order to reduce
the optical resolution to the required final output resolution. However, there is ample
empirical evidence which indicates that decimation aliasing at/near the Nyquist
frequency can be substantially avoided with the correct algorithm, while maintaining
sharpness and detail rendition. On the other hand, aliasing due to upscaling is not
acceptable at any frequency up to and including the Nyquist frequency, because it implies
that the true optical resolution is lower than the required 500 ppi or 1000 ppi. More
details of this quantitative alias detection technique can be found in the MTF document7
which is on the IQS Test Tools CD.

The DFT-based, quantitative assessment of aliasing is dependent on the threshold levels

chosen, which have been judiciously chosen in sinemtf, based on experimentation. It is
useful, however, to apply other alias detection techniques, in order to obtain as complete
a picture of aliasing as is possible. As it happens, one of the benefits of a sine wave
target (or bar target) is that it readily lends itself to showing the effects of aliasing at
specific frequencies; e.g., visible banding is one tell-tale sign of aliasing; for more detail,
see illustrations in the Rescale document8 which is on the IQS Test Tools CD. A more
comprehensive assessment of the magnitude and potential impact of aliasing is therefore
achieved by the synergistic combination of the quantitative, DFT approach and the
qualitative, visual assessment approach.

2.3.4 Requirements Compliance

1) The MTF, if applicable, shall meet the minimum modulation requirements given in
Table 2-2. The CTF, if applicable, shall meet the minimum modulation requirements as
defined in equation 2-1 or equation 2-2 (whichever applies). The MTF or CTF
modulation requirements shall be met for both vertical and horizontal target pattern
orientations, and for the target locations as described in Section 2.3.3.

7“Computer Program to Determine the Sine Wave MTF of Imaging Devices”, N.B.Nill, D.J.Braunegg,
B.R.Paine, MITRE Corp. Technical Report, MTR-96B025, June 1996; Section 2.8, “Detection of Aliasing”

8“Rescaling Digital Fingerprints: Techniques and Image Quality Effects”, D.J.Braunegg, R.D.Forkert,
N.B.Nill, MITRE Corp. Technical Report, MTR-95B061, June 1995.

26
2) None of the MTF or CTF modulation values measured at specification spatial
frequencies shall exceed 1.05.

3) When applying a suitable quantitative alias detection metric, supported by visual

observation, the aliasing requirement is met when no substantial aliasing-due-to-
decimation is detected at any frequency less than 70% of the required output resolution,
and no aliasing-due-to-upscaling is detected at any frequency less than or equal to the
required output resolution.

2.4 SIGNAL TO NOISE RATIO TEST

2.4.1 Requirements

The white signal-to-noise ratio and black signal-to-noise ratio shall each be greater than
or equal to 125.0, in at least 97% of respective cases, within each printblock
measurement area.

2.4.2 Background

The signal is defined as the difference between the average output gray-levels obtained
from scans of a uniform low reflectance and a uniform high reflectance target, measuring
the average values over independent 0.25 by 0.25 inch areas (“quarter-inch areas”) within
each printblock area. The noise is defined as the standard deviation of the gray-levels in
each quarter-inch measurement area. Therefore, for each high reflectance, low
reflectance quarter-inch image pair, there are two SNR values, one using the high
reflectance standard deviation and one using the low reflectance standard deviation. The
scanner must be set up such that the average image gray-level of the high reflectance
target is below 255 or high clipping level, whichever is lower, and the average image
gray-level of the low reflectance target is above 0 or low clipping level, whichever is
higher. Note that in this method of measuring SNR, no attempt is made to isolate
different sources of noise or separately measure different types of noise; the computed
noise represents all noise types and sources taken together.

2.4.3 Target

For a paper scanner, two uniform, neutral gray targets with matte reflectance on paper or
mylar base are used. One target has high reflectance, denoted as the white target, such as
Munsell model N9 (79% reflectance). The other target has low reflectance, denoted as
the black target, such as Munsell model N3 (7% reflectance). [See Appendix B for target
manufacturers.]

For a card scanner, the white and black targets can be mounted together, covering a 5 x 8
inch area of a single, 8 x 8 inch card stock, as depicted in Figure 2-9. Alternatively, the

27
white and black targets can be mounted on separate 8 x 8 inch card stocks, each covering
a 5 x 8 inch measurement area.

For live scanner testing a different target substrate may be required; e.g., flashed
photographic emulsion on mylar base. Alternatively, it may be necessary to separately
capture equivalent white and black images with a blank livescan platen (no actual target),
such as by inserting a neutral density filter in the optical path or by adjusting the detector
integration time. In either case, the target or pseudo-target should cover the entire
livescan image capture area.

Note that the input/output linearity established for the test environment needs to extend to
these white and black target reflectance values.

2.4.4 Test Procedures

1) The scanner is adjusted so that the average output gray-level of the white target is at
least 4.0 gray-levels below the maximum gray-level, and the average output gray-level of
the black target is at least 4.0 gray-levels above the minimum gray-level. For a system
that captures/outputs the full 8 bpp, 256 gray-level range, the maximum and minimum
gray-levels are 255 and 0, respectively, in which case the white average must be less than
or equal to 251.0 and the black average must be greater than or equal to 4.0. However, if
the scanner system is setup such that some gray-levels cannot occur, then the maximum
or minimum value must be adjusted accordingly. For example, if gray-levels 253, 254,
255 cannot occur because they are always clipped-out of an image (for whatever reason),
then the maximum gray-level would be 252 and the white average would need to be less
than or equal to 248.0.

2) If using the composite target illustrated in Figure 2-9, then the composite target is
scanned, rotated 180 degrees, and scanned again. If using separate black and white
targets, each target is scanned. In either approach, the two images result in both a white
and black image segment on both the left and right sides of the total image area.

3) The locations of all independent quarter-inch windows that fit within each printblock
area are identified. For a card scanner this produces 6 rows and 6 columns of quarter-
inch windows for each rollblock, for a total of 36 windows per rollblock, 21 windows for
each plain thumb block, and 84 windows for each 4-finger plain block. For a live
scanner, all independent quarter-inch windows that fit within the capture area are
identified.

4) The average, x, and the standard deviation, s, for each of the quarter-inch windows in
the white and black target scans are computed. The SNR value for each quarter-inch

28
black/white window pair9 is then computed using swhite and sblack, according to the
formulas:

é xwhite - x black ù
SNR white =
êë s white úû

éx - xblack ù
SNR black = ê white
ë s black úû

5) Notes:

- The SNR requirement can be verified with the snr software, which is on the IQS Test
Tools CD. [Snr avoids the area around the white/black target join line shown in
Figure 2-9.]

- If, after careful cleaning of the target and scanner platen, a small quantity of
measurement samples still contain residual artifacts, such as dust, pinholes, scratches,
or smudges on the target, then these samples may be discounted from the final test
sample size.

2.4.5 Requirements Compliance

The requirement is complied with if SNRwhite and SNRblack are each greater than or
equal to 125 in at least 97 percent of the respective cases, in each of the fourteen print
blocks for a card scanner, and over the total image capture area for a live scanner.

9 A “window pair” is a quarter-inch black window and quarter-inch white window, alternately occupying
the same location in the total capture area.

29
 8″!

LEAVE BLANK! TYPE OR PRINT ALL INFORMATION IN BLACK! FBI! LEAVE BLANK!
LAST NAME NAM! FIRST NAME MIDDLE NAME !
APPLICANT!

SIGNATURE OF PERSON FINGERPRINTED! ALIASES AKA! O

R!
I!
RESIDENCE OF PERSON FINGERPRINTED! DATE OF BIRTH DOB!
Month Day Year!

3″!
CITIZENSHIP CTZ! SEX! RACE! HGT.! WGT.! EYES! HAIR! PLACE OF BIRTH POB!
DATE! SIGNATURE OF OFFICIAL TAKING FINGERPRINTS!

YOUR NO. OCA!

LEAVE BLANK!
EMPLOYER AND ADDRESS!
FBI NO. FBI!

ARMED FORCES NO. MNU! CLASS_______________________________________________!

!
REASON FINGERPRINTED!
SOCIAL SECURITY NO. SOC!
!
REF._________________________________________________!
MISCELLANEOUS NO. MNU!

1.6 ! 1.5 !

1. R THUMB !
79%
2. R. INDEX! 3. R. MIDDLE! 4. R. RING!
7% 5. R. LITTLE!

reflectance reflectance

5″!
6. L THUMB ! 7. L. INDEX! 8. L. MIDDLE! 9. L. RING! 10. L. LITTLE!

0.8 !
3.2 !
1.9 !
LEFT FOUR FINGERS TAKEN SIMULTANEOUSLY! L. THUMB! R. THUMB! RIGHT FOUR FINGERS TAKEN SIMULTANEOUSLY!

4″!
Figure 2-9. Composite of Two Uniform Gray Targets

30
2.5 GRAY-LEVEL UNIFORMITY

2.5.1 Requirements

#1 - Adjacent Row, Column Uniformity:

At least 99% of the average gray-levels between every two adjacent quarter-inch long
rows and 99% between every two adjacent quarter-inch long columns, within each
imaged printblock area, shall not differ by more than 1.0 gray-levels when scanning a
uniform low reflectance target, and shall not differ by more than 2.0 gray-levels when
scanning a uniform high reflectance target.

#2 - Pixel to Pixel Uniformity:

For at least 99.9% of all pixels within every independent 0.25 by 0.25 inch area located
within each imaged printblock area, no individual pixel's gray-level shall vary from the
average by more than 22.0 gray-levels, when scanning a uniform high reflectance target,
and shall not vary from the average by more than 8.0 gray-levels, when scanning a
uniform low reflectance target.

#3 - Small Area Uniformity:

For every two independent 0.25 by 0.25 inch areas located within each imaged printblock
area, the average gray-levels of the two areas shall not differ by more than 12.0 gray-
levels when scanning a uniform high reflectance target, and shall not differ by more than
3.0 gray-levels when scanning a uniform low reflectance target.

2.5.2 Target

The targets are the same as used for SNR assessment, see Section 2.4.3.

2.5.3 Test Procedures

2.5.3.1 Preparation

The target scans are the same scans used in SNR assessment, see Section 2.4.4.

2.5.3.2 Adjacent Row, Column Uniformity Test Procedure

For each of the fourteen print block areas of the white target and black target, the average
pixel values of individual 0.25 inch long horizontal row segments and individual 0.25
inch long vertical column segments are computed (avoiding column or row segments
near the black/white target join line shown in Figure 2-9). For a given image (black or
white), the magnitude of the difference between the average values of every two adjacent
row segments and every two adjacent column segments are computed and compared to
the requirement. Figure 2-10 illustrates the locations of some of the horizontal segments
in a rolled impression print block image.

31
 1, 26 1,151 1,651
R=1, C=1
a b c d e f
h i j k l m
o p q r s t

125 pixels
per row segment

Labels a-u represent the average gray values of some of the 125 pixel
row segments; the magnitude differences of all adjacent row
segments are determined: |a-h|, |h-o|, |b-i|, |i-p|, etc.

1.5″
.
.
.

750, 26

1.6″

Figure 2-10. Example Measurements for Adjacent Row Uniformity Assessment

(125 Pixel Long Row Segments in Rollblock, 500 ppi Scanner)

2.5.3.3 Pixel to Pixel Uniformity Test Procedure

The locations of all independent quarter-inch windows that fit within each printblock area
are identified. For example, for a ten-print card scanner this produces 6 rows and 6
columns of quarter-inch windows for each rollblock, for a total of 36 windows per
rollblock, 21 windows for each plain thumb block, and 84 windows for each 4-finger
plain block. For a live scanner, all independent quarter-inch windows that fit within the
total capture area are identified.

The average gray-level for each quarter-inch window, rounded to the nearest whole
number (nearest integer value), is computed.

32
The absolute value of the difference between the average of a given quarter-inch window
and each of the individual pixel values within that window is computed. For a 500 ppi
scanner, there are 125 x 125 = 15,625 test values in each quarter-inch window.

2.5.3.4 Small Area Uniformity Test Procedure

The average gray-level for each quarter-inch window defined in Section 2.5.3.3 is
recomputed, this time without rounding, producing a floating point number.

For a given target scan, the absolute value of the difference between the averages is
computed for every possible pair of quarter-inch windows within each of the fourteen
print blocks; or, for a live scanner, over the total image capture area. This difference is
denoted as Awhite for the white target cases and Ablack for the black target cases. Awhite
and Ablack each have 630 values for each rollblock (from 36 quarter-inch windows), 210
values for each plain thumb block (from 21 quarter-inch windows), and 3486 values for
each plain four-finger block (from 84 quarter-inch windows).

Notes:
- The gray-level uniformity requirements can be verified with the snr software
which is on the IQS Test Tools CD. [Snr avoids the area around the white/black
target join line shown in Figure 2-9.]

- If, after careful cleaning of the target and scanner platen, a small quantity of
measurement samples still contain residual artifacts, such as dust, pinholes,
scratches, or smudges on the target, then these samples may be discounted from
the final test sample size.

2.5.4 Requirements Compliance

The adjacent row, column uniformity requirements are complied with if, for each
printblock area, at least 99% of the differences between the average values of adjacent
quarter-inch long rows, and at least 99% of the differences between the average values of
adjacent quarter-inch long columns, do not differ by more than 1.0 gray-levels when
scanning a uniform low reflectance target, and do not differ by more than 2.0 gray-levels
when scanning a uniform high reflectance target.

The pixel to pixel uniformity requirement is complied with if at least 99.9 percent of all
of the pixels in each of the quarter-inch windows, within each imaged printblock area, is
within 22 gray-levels of the window’s mean gray-level for the white target image, and is
within 8 gray-levels of the window’s mean gray-level for the black target image.

33
The small area uniformity requirement is complied with if Awhite ≤ 12.0, for all of the
Awhite values within each imaged printblock area, and if Ablack ≤ 3.0, for all of the Ablack
values within each imaged printblock area.

[For a livescanner, the “printblock area” corresponds to the total image capture area.]

2.6 FINGERPRINT IMAGE QUALITY

2.6.1 Requirements

Fingerprint Gray Range:

At least 80% of the captured individual fingerprint images shall have a gray-scale
dynamic range of at least 200 gray-levels, and at least 99% shall have a dynamic range of
at least 128 gray-levels.

Fingerprint Artifacts and Anomalies:

Artifacts or anomalies detected on the fingerprint images, due to the scanner or image
processing, shall not be significant enough to adversely impact support to the functions of
conclusive fingerprint comparisons (identification or non-identification decision),
fingerprint classification, automatic feature detection, or overall Automated Fingerprint
Identification System (AFIS) search reliability.

Fingerprint Sharpness and Detail Rendition:

The sharpness and detail rendition of the fingerprint images, due to the scanner or image
processing, shall be high enough to support the fingerprint functions of conclusive
fingerprint comparisons (identification or non-identification decision), fingerprint
classification, automatic feature detection, and overall Automated Fingerprint
Identification System (AFIS) search reliability.

2.6.2 Target

A card scanner is tested with an FBI-supplied set of 8 x 8 inch, ten-print, inked

fingerprint cards. These cards represent a set of stratified random samples drawn from
the FBI's Fingerprint Card Master File (FCMF). This test card set consists of three strata:
light-inked cards, medium-inked cards, and dark-inked cards. Appendix D describes the
construction of a set of stratified card samples.

For 500 ppi testing, at least the 5 x 8 inch, 14 printblocks area should be scanned; for
1000 ppi testing it is sufficient to scan the 3 x 8 inch, 10 rollblocks area. In either case
the entire 8 x 8 inch card may be scanned.

This card set includes a number of cards with difficult to handle properties; e.g., tears,
holes, staples, glued-on photos, lamination, etc. If the card scanner is to be certified for

34
use with an Automatic Document Feeder (ADF), then an attempt should be made to feed
all cards through the ADF for scanning. If some of the difficult to handle cards cannot be
fed through the ADF, they should be scanned in whatever manual mode is available. It is
expected that the vendor would then establish an exceptions handling procedure and this
procedure would be conveyed to the user for operational employment.

For a live scanner operating at 500 ppi or 1000 ppi, the vendor seeking certification is
responsible for supplying the following sets of livescans:

• For a standard roll and plain finger live scanner: capture a complete set of
fingerprints from each of 10 subjects; i.e., 10 rolls (all 5 fingers from each hand), 2
plain thumb impressions, and 2 plain 4-finger impressions.

• For a palm scanner component of a live scan system: capture left and right palms
from each of 10 subjects.

• For an Identification Flats live scanner: capture left and right 4-finger plain
impressions and dual thumb plain impressions from each of 10 subjects.

For Fast Track testing, the vendor normally supplies 10 cardscans or livescan sets from 5
subjects.

The test cardscans or test livescans must be supplied in an uncompressed format such as
raw, TIFF, or PGM format (not in NIST or EBTS format). Furthermore, these
uncompressed images cannot have been previously compressed/decompressed via a lossy
compression, such as WSQ.

2.6.3 Test Procedure

2.6.3.1 Fingerprint Gray Range

1) Whether scanning ten-print cards or livescans, the same basic methodology is used;
i.e., a single subimage is first defined within the scanned image of each fingerprint block.
This subimage is sized and positioned such that it includes a substantial part of the
fingerprint, while excluding format lines, box lines, printed/handwritten text, and most of
the white background. Figure 2-11 gives some examples of acceptable subimage
sizing/positioning under several different conditions in a ten-print card. In some cases, a
printblock-centered subimage of constant size (for each of the 4 printblock types) could
fulfill the capture requirements.

35
 Figure 2-11. Example Measurement Boxes for Gray Range Assessment
(Width/Height Ratio of Each Box is the Same as the Printblock Width/Height Ratio).

2) The gray range is computed for each of the subimages defined in step (1). The gray range
is equal to the total number of gray-levels in the subimage which contain signal, where a
gray-level bin is counted as containing signal if it contains at least a minimum number of
pixels.

Background:
- Since a subimage contains hundreds of thousands to millions of pixels, the
expectation is that if a given gray-level bin contains signal, then it would be
populated by more than just a few pixels, since all signal pixels are spread out
between no more than 256 gray-level bins. Therefore, if a gray-level bin contains
very few pixels it is probably just noise; e.g., dark current, crosstalk, or amplifier
noise. A threshold value of 5 pixels can be used to separate gray-level bins populated
only by noise, from bins populated by signal (+ noise).

- The 8 bpp quantization of the gray-scale values for very low contrast fingerprints
needs to more optimally represent the reduced gray-scale range of such
fingerprints, which may require adaptive processing. The intent of such

36
 processing would be to overcome excessively low contrast images without adding
false detail.

- Note that the definition of gray range in this section is not, in general, equal to the
simple difference between maximum gray-level and minimum gray-level.

3) For card scans, the subimages from the cards corresponding to each of the three strata
are grouped together and a table such as the example case given in Table 2-6 is generated,
from the computed dynamic ranges of the subimages in each stratum.

Table 2-6. Example 50 Card Test Results

Stratum Number of Total Number of Number of Stratum

Cards Images Images with Images in Weight
1 to 127 200-256
Gray-Levels Gray-Levels
1 (light) 20 200 1 196 0.0314
2 (medium) 20 200 0 200 0.9628
3 (dark) 10 100 0 100 0.0058

The values in Table 2-6 are used to calculate the strata-weighted results, by first
computing for each stratum:

(stratum weight) x (#images in gray range being calculated) / (total images in stratum)

and then summing over the three strata. Using the example values from Table 2-6, this
procedure results in,

This analysis, which is described in more detail in Appendix D, is implemented in the

grayfinger software, which is on the IQS Test Tools CD.

37
2.6.3.2 Fingerprint Artifacts and Anomalies

The fingerprint images will be examined to determine the presence of artifacts or

anomalies due to the scanner or image processing. Artifacts and anomalies such as the
following non-inclusive list may be investigated:

- jitter noise effects

- sharp truncations in average gray-level between adjacent printblocks
- gaps in the gray-level histograms; i.e., zero pixels in intermediate gray-levels, or
clipping to less than 256 possible gray-levels
- imaging detector butt joints
- noise streaks
- card bleed-through
- excessive gray-level saturation

Due to the varied nature of potential artifacts and anomalies, there is no single test tool or
analysis method that can be applied to all cases. As a first step, the fingerprint images
are displayed and inspected for any obvious problems. If a significant artifact or
anomaly is visually detected, then the next stage of analysis is to decide whether it is due
to the input fingerprint (inked or live), or is due to the scanner/image processing
combination; if it is due to the scanner/image processing combination, then appropriate
image analysis techniques are applied to measure and quantify it. Other problems may
not be obvious from visual assessment, for example, the grayfinger software on the IQS
Test Tools CD detects and reports-out the existence of gray-level clipping, which may
not be visually discernable.

2.6.3.3 Fingerprint Sharpness and Detail Rendition

The fingerprint images are displayed and inspected to determine their visual sharpness
and detail rendition; e.g., by comparing a given image to others in the set, or comparing a
given image to other compatible sets of images. Visual assessment can, for example,
detect images that are significantly out-of-focus.

Assessments of sharpness and detail rendition can be quantified by applying a suitable

objective quality metric. Although no definitive fingerprint quality metric is identified
here, the following is a partial list of metrics which may be useful in this application. In
applying any objective quality metric in this testing venue, it is important to keep in mind
that the goal is to assess scanner quality, not input finger quality.

• The NIST nfiq fingerprint image quality software and documentation available for
download at:
http://www.itl.nist.gov/iad/894.03/nigos/nigos.html

38
 • The MITRE iqf fingerprint image quality software and documentation available
for download at:
http://www.mitre.org/tech/mtf

2.6.4 Requirements Compliance

1) The fingerprint gray range requirement is met if at least 99 percent of the strata-
weighted set of cardscan subimages, or unweighted subimages for livescans, have at least
128 gray-levels, and at least 80 percent of the same set of subimages have at least 200
gray-levels.

2) The artifacts and anomalies requirement is met if detected artifacts or anomalies, due
to the scanner or image processing, are not significant enough to adversely impact
support to the functions of conclusive fingerprint comparisons (identification or non-
identification decision), fingerprint classification, automatic feature detection, or overall
Automated Fingerprint Identification System (AFIS) search reliability.

3) The sharpness and detail rendition requirement is met if the sharpness and detail
rendition of the fingerprint images, due to the scanner or image processing, is sufficient
to support the functions of conclusive fingerprint comparisons (identification or non-
identification decision), fingerprint classification, automatic feature detection, and overall
Automated Fingerprint Identification System (AFIS) search reliability.

39
(blank page)

40
 SECTION 3

IDENTIFICATION FLATS SCANNER

3.1 Requirements

Inclusive:
All requirements stated in Section 2 apply.

Capture Protocol:
The system shall provide a simple capture protocol so that the inexperienced user
consistently captures high quality fingerprints.

Verifiable Finger Sequence Data:

The fingers shall be captured in a way that eliminates the possibility of error in the finger
numbers.

3.2 Background

Traditional fingerprint sets contain both rolled and plain fingerprint images. The rolled
impressions support the search processing and identification functions and the plain
impressions are used primarily for sequence verification. Fingerprint systems designed
for Identification Flats civilian background checks capture a single set of plain
impressions. This single set of plain impressions must support finger sequence
verification, search processing, and identification.

Image quality has historically been a challenge for civil background checks. Some
programs require a large number of relatively low volume capture sites, which makes
training difficult. A key requirement for identification flats scanners is to reduce the
training requirement so that inexperienced users consistently capture quality fingerprint
images.

3.3 Test Procedures

The test procedures described in Sections 2.1 to 2.5, which utilize measurements on
images of deterministic targets, are performed.

The test procedures described in Section 2.6 “Fingerprint Image Quality” are performed
by assessments on the left and right 4-finger plain impressions and dual thumb plain
impressions from each of 10 subjects.

The system will be evaluated for its ability to produce a very small rate of failure-to-
enroll in an operational setting.

41
The fingerprint system capture protocol will be evaluated for its ability to capture
verifiable finger sequence data.

3.4 Requirements Compliance

Inclusive:
All requirements stated in Section 2 shall be met.

Capture Protocol:
The requirement is met if the system has a minimum capture area of 3.2 inches (width)
by 3.0 inches (height) and can capture 4 fingers simultaneously in an upright position.
Other capture approaches will require specific testing and documentation.

Verifiable Finger Sequence Data:

The requirement is met if the system has a minimum capture area of 3.2 inches (width)
by 3.0 inches (height) and can capture the left four fingers simultaneously, the right four
fingers simultaneously, and the two thumbs simultaneously (4-4-2) in an upright position.
Other capture approaches will require specific testing and documentation.

42
 SECTION 4

FINGERPRINT PRINTER

In this section, the term “printer” refers to the combination of a physical printer,
fingerprint printing algorithm, and print medium, which shall be paper or card stock.
A required printer resolution is 500 ppi

Verification of the printer performance requirements is accomplished by evaluating the

printer's life-size output prints of an FBI-designated test set of digitized FingerPrints
(FP), and evaluating an FBI-designated digital test TarGeT (TGT). One component of
the FP test set is printed in 10-print card format. Version A8 of the TGT is shown in
Figure 4-1.

Requirements compliance verification is performed by a combination of visual

assessments of the FP and TGT prints (aided by visual instruments such as magnifiers),
and computer-aided assessments of scans of the FP and TGT prints. With respect to
those requirements that depend on assessments of print scans for compliance verification,
the scan resolution is expected to be twice the required print resolution; e.g., a 500 ppi
print would be scanned at 1000 ppi, and the scanner would be setup in a calibrated linear
input/output, grayscale reflectance capture mode.

For a non-laser printer, e.g., inkjet printer, it may be required to stipulate additional
parameters that would become part of the certification listing, such as type of paper used,
type/color of ink cartridges used, or user-controlled features of the printer which could
affect image quality.

4.1 Spatial Frequency Response

4.1.1 Requirements

The printer shall provide sufficient spatial frequency response to support visually
resolving the required printer resolution, in orthogonal directions on the print.

When an appropriate method for measuring the spatial frequency response curve is
applied, that curve shall meet or exceed a minimally acceptable level, in orthogonal
directions on the print.

43
 5-Bar & 3-bar targets 500 ppi scanned, inked
low & high contrast fingerprint
1.1 cy/mm Ronchi Bars

45 degree line

Sine Waves

components
description

V&H
sharp edges

A8

64 Gray Steps 256 Gray Steps

Frame Line

Figure 4-1. Printer Test Target (TGT)

(Version A8 Designed for Printing at 500 ppi, Outer Frame is 4000 by 2500 Pixels)

4.1.2 Test Procedures

When the TGT and the ten-print card format component of the FP test set are printed life-
size, each print contains black/white parallel bar patterns corresponding to the spatial
frequency of the required print resolution. For example, a fingerprint captured at 500 ppi
has a limiting frequency of 250 cycles per inch and the corresponding FP and TGT prints
would each contain 250 cycles per inch bar patterns.

The bar patterns at the limiting frequency are observed and a determination of
resolvability is made with the aid of a magnifier, typically in the 4X to 10X magnification

44
range. Resolvability of a given bar pattern is a binary decision; i.e., it is either judged to
be visually resolved or not resolved10.

A bar pattern at the limiting frequency represents a single point on the printer’s spatial
frequency response curve. The entire spatial frequency response curve may be measured
by scanning the print of an appropriate target and performing appropriate computer-aided
assessment on the scan.

4.1.3 Requirements Compliance

At least half of all high contrast bar patterns11 in each direction (vertical and horizontal),
and at the spatial frequency corresponding to the required printer resolution, shall be
visually resolved on the TGT print.

The vertical and horizontal direction bar patterns (“bar chart”) in the FP prints shall be
visually resolved.

When an appropriate target and computational method is used for measuring the spatial
frequency response curve of the printer, the measured curve shall meet or exceed a
minimally acceptable response curve.

4.2 Gray-levels

4.2.1 Requirements

At least 16 gray-levels shall be visually distinguishable on the print.

4.2.2 Test Procedures

The TGT is printed lifesize. TGT version A8 contains a 64 gray-step tablet, spanning the
range from black (gray=0) to white (gray=252), with 3 gray levels between adjacent
steps. Each gray-level step is large enough on the print to be individually distinguishable
(~1/3 x 1/3 inch).

The observer counts the number of individually distinguishable steps on the print,
utilizing the following guidelines:
- place a blank, white sheet of paper underneath the print

10Guidance on judging resolvability can be found in the international standard: Micrographics - ISO
Resolution Test Chart No. 2 - Description and Use, ANSI/AIIM MS51-1991 and ISO 3334:1991.

11 One “bar pattern” is a set of equal-width, adjacent, parallel bars corresponding to a specific spatial
frequency.

45
 - use adequate, non-glare room lighting
- in some cases a low power magnifier (~2X)12 may aid assessment
- a given visual gray-level is only counted once, even though it may appear in more
than one step
- do not count a visual gray level that is out-of-sequence
- count as a single gray level, all adjacent steps (two or more) that appear to have the
same gray level
- concentrate on the central area of a step for assessment; this is particularly
necessary if the print exhibits noticeable edge enhancement.

4.2.3 Requirements Compliance

At least 16 gray-levels shall be visually distinguishable on the print.

4.3 Dynamic Range

4.3.1 Requirements

The printer shall have the capability to print an input digital image gray range of at least
130, excluding print black saturation and print white saturation.

4.3.2 Test Procedures

The print of a digital step tablet is scanned; e.g., the 64 step tablet in TGT version A8.
Each pixel’s output gray-level value is converted to the corresponding print reflectance
value, and the average print reflectance value within each step is computed. A plot of
step average print reflectance versus input digital step tablet gray level is constructed, as
illustrated in Figure 4-2.
[The scanner output gray-level to print reflectance conversion must first be established by
generating the scanner’s input/output curve using a calibrated step tablet. The dynamic
range can then be verified by plotting the output of the step64 software, which is on the
IQS Test Tools CD.]

4.3.3 Requirements Compliance

Excluding any saturation on the low end (print black reflectance) and high end (print
white reflectance), the remaining print reflectance range shall correspond to a target gray
range of at least 130 gray levels. The following is a deterministic method used for
excluding saturation effects.

12A 2X magnifying power means that the object appears to the eye to be 2 times larger than it would
appear to the unaided eye at normal reading distance.

46
Let delta be the difference in Reflectance values between the highest and lowest
saturation values of an ideal S-shaped curve. The low threshold of the dynamic range is
defined as being 5% of delta above the low saturation value of the S-shaped curve. The
high threshold of the dynamic range is defined as being 5% of delta below the high
saturation value of the S-shaped curve. The difference between the high threshold and
the low threshold values is the dynamic range. Departure from the ideal S-shaped curve
(bumps and dips) will be assessed on a case-by-case basis.

Print
1
Reflectance
0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1
gray range = 165
0
0 50 100 150 200 250
Input Digital Target Gray Level
(a)
1
Reflectance
(Print)
0.9

0.8
-5% D

0.7

0.6
Dynamic Range
0.5

0.4
D

0.3

0.2

0.1
+5% D

0
0 50 100 150 200 250
Gray Levels (Target)

47
 (b)
Figure 4-2. Print Dynamic Range Measurement
4.4 Geometric Accuracy and Print Scale

4.4.1 Requirements

When printing a digital bar target containing parallel bars, then the absolute value of the
difference (D), between the measured distance (Y) across the parallel bars on the print,
and the correct distance (X) on the print, shall not exceed the values for D given in Table
4-1, for at least 97% of the tested “short distance” and “medium distance” cases in each
direction (vertical and horizontal).

The average of all “medium distance” test cases, in each direction, shall not exceed the
corresponding values for D given in Table 4-1.

The average of all “long distance” test cases, in each direction, shall not exceed the
corresponding values for D given in Table 4-1.

Straight target lines printed parallel to, or at a 45-degree angle to, the paper or card edges,
shall be straight on the print, with no significant waviness, bow, or staircasing.

Table 4-1 Geometric Accuracy Requirements

Distance Error (D) Distance Range (X) Comment

D ≤ 0.001 0.00 < X ≤ 0.07 short distance
D ≤ 0.015X 0.07 < X ≤ 1.50 medium distance
D ≤ 0.010X 4.75 < X ≤ 8.00 long distance

Table Note: D = |Y-X|

X = correct distance = digital target pixels / required print resolution
Y = measured distance on print
D, X, Y are in inches

4.4.2 Test Procedures

The TGT is printed lifesize. Target version A8, when printed at 500 ppi, contains a 7.1
inch long segment of parallel vertical bars and a 4.7 inch long segment of parallel
horizontal bars on the print,13 where the bar frequency is: 500 / (25.4*18) = 1.0936

13 These bar patterns actually contain two components, either one of which could be used for requirement
compliance testing: (1) parallel bars with equal bar and space widths at 1.0936 cy/mm, known as a Ronchi
target, and (2) narrow, parallel bars with wide spaces, also at 1.0936 cy/mm. The geo software operates on
the equal bar and space Ronchi bar pattern.

48
cy/mm. The TGT also contains frames lines around the border and lines at a 45 degree
angle to the target edges.

Assessment of the short distance and medium distance geometric accuracy, and medium
distance print scale, requires that the print be scanned at 1000 ppi. When the print is
oriented such that the bar lengths in the bar segment to be measured are in-line with the
scanner’s direction of scan, as illustrated in Figure 4-3, then scanner-induced
perturbations are minimized (this requires 2 scans to capture the two orthogonal Ronchi
segments, with a 90 degree rotation of the print on the scanner bed between scans). The
scanner’s output digital image can then be input to the geo software, supported by
creategeofile and viewgeo software, which are on the IQS Test Tools CD. See Appendix
C for a detailed description of the computations performed in geo.

Assessment of the long distance print scale can be performed manually, measuring the
distance between the 8 x 5 inch frame lines and/or 7.75 x 4.75 inch frame lines (left/right
and top/bottom frame lines), with the aid of a precision ruler and magnifier. The % scale
error in the print can be computed from:

100 | printwidth - targetwidth |

% scale error =
targetwidth

where,
number of pixels between 2 parallel, vertical frame lines
targetwidth =
target ppi

printwidth = measured width on print between the same 2 parallel, vertical frame
lines that are used in calculating the targetwidth

Line straightness can be assessed manually, by inspecting the vertical and horizontal
framelines and 45 degree lines, with the aid of a straight-edge and magnifier.

Ronchi bars

scan
direction scanner's linear sensor array

49
Figure 4-3. Correct Orientation of Print Ronchi Bars with Respect to Scanning Array

Background:

For the short and medium distance measurements and 500 ppi print resolution, a digital
bar target with a period of 18 pixels is used, which corresponds to a bar frequency of
1.0936 cy/mm on the print, when printed life-size. The measured distance on the print
can be obtained by scanning the print and applying computer-assisted assessment on the
resulting digital image. The allowable short distance and medium distance errors given
in Table 4-2, take into account the geometric errors inherent in a good quality scanner.

Medium distance print scale is measured from the bar target, where the ±1.5% allowable
average error corresponds to an allowable print ppi range of 492.5 to 507.5 ppi. Due to
the averaging and balancing-out of random, plus/minus medium distance scale errors, it
could reasonably be expected that the long distance scale error would be less than the
±1.5% error allowed for medium distance. Therefore, long distance print scale, measured
between the vertical and horizontal frame lines of the TGT, has an allowable ±1.0%
average error in each direction.

Table 4-2. Geometric Accuracy Tests

Distance Measurement Correct Distance Directional Requirement Met if:

(inches)
Short distance (1-bar) 0.0360 ≥ 97% in range: 0.0350 to 0.0370
Medium distance (6-bar) 0.2160 ≥ 97% in range: 0.21276 to 0.21924

avg in range: 0.21276 to 0.21924

Long distance avg in range:
(frame lines) 8.000 7.9200 to 8.0800
5.000 4.9500 to 5.0500
7.7500 7.6725 to 7.8275
4.7500 4.7025 to 4.7975

Note: 500 ppi print with 1.0936 cy/mm bars.

4.4.3 Requirements Compliance

The short distance and medium distance geometric accuracy requirements are complied
with if at least 97 percent of the tested cases in each direction (vertical and horizontal), in

50
each distance range, are within the minimum and maximum distance limits defined in
Table 4-2.

The medium distance print scale requirement is complied with if the average distance in
each direction (vertical and horizontal) is within 1.5% of the correct distance, which is
equivalent to the average ppi being between 492.5 and 507.5.

The long distance print scale requirement is complied with if the average distance in each
direction (vertical and horizontal) is within 1.0% of the correct distance.

The line straightness requirement is complied with if inspection of the long vertical and
horizontal frame lines, and the 45 degree lines, indicates that these lines are straight, with
no significant waviness, bow, or staircasing.

4.5 Noise

4.5.1 Requirements

For a required printer resolution of 500 ppi, the noise magnitude shall be less than 0.120 at
all print reflectance levels, when noise magnitude is defined as the standard deviation of
print reflectance values within an area on the print corresponding to a constant gray level
on the input digital target. [Print reflectance measured in fractional units: 0.0 to 1.0 range.]

4.5.2 Test Procedures

The TGT is printed lifesize and noise assessment is applied to the 64-gray step tablet as
follows:

1) A calibrated reflection step tablet14 is scanned in a flatbed scanner at 1000 ppi and the
scanner is adjusted to obtain a linear or near-linear relation between target reflectance and
scanner output gray-level; the curve of best fit is determined for this relation, with target
reflectance as the independent variable.

2) The print of the TGT is scanned at 1000 ppi on the same scanner, setup in its
calibrated mode.

3) The standard deviation in reflectance units is computed for each of the 64 steps, using
the scanner output digital image and the scanner calibration curve. [The step64 software
on the IQS Test Tools CD performs these computations.]

14Reflection step tablets, where each step is calibrated in density or reflection units, are commercially
available - see Appendix B.

51
4.5.3 Requirements Compliance

Each of the 64 computed standard deviations of print reflectance values shall be less than
0.120.

4.6 Print Polarity and Color

4.6.1 Requirements

The printed fingerprints shall appear as dark gray-to-black ridges on a light gray-to-white
background.

4.6.2 Requirements Compliance

The printed fingerprints are visually inspected for compliance with the requirement.

4.7 Print Permanence

4.7.1 Requirements

The printed fingerprints shall not smear or smudge with normal handling.

4.7.2 Requirements Compliance

Compliance testing may consist of: (1) using moderate finger pressure to rub a fingerprint
area of the print and then establish whether or not any visually apparent smearing or
smudging of the fingerprint has occurred, and (2) adding a drop of water to a fingerprint
area of the print, allowing it to dry undisturbed, and then establish whether or not any
visually apparent smearing or smudging of the fingerprint has occurred.

Prints on paper or card stock from a standard laser printer15 meet the print permanence
requirement without testing, as do prints on paper or card stock from any other type of
printer that has print permanence characteristics equivalent to or better than a laser printer.

4.8 Print Stability

4.8.1 Requirements

15 “Standard laser print system” here refers to a type of print system in which a laser beam “draws” an
electrostatic image of an input signal onto a drum. Black toner (typically dry powder) is then transferred to
the charged areas of the drum, which then transfers the toner onto paper, where it is fused by heat, creating
a black/white/gray image.

52
Both the printed fingerprints and the card stock or paper on which they are printed shall
retain their visually neutral (black, white, gray) color over time.

4.8.2 Requirements Compliance

Prints on paper or card stock from a standard laser printer meet the print stability
requirement, as do prints on paper or card stock from any other type of printer that has
stability characteristics equivalent to or better than a laser printer.

4.9 Hazardous Materials

4.9.1 Requirements

The prints shall not produce any health hazard as a result of handling. They shall not
produce any noxious, annoying, or unpleasant odors when accumulated in large numbers
and handled in areas having limited ventilation.

4.9.2 Requirements Compliance

Prints on paper or card stock from a standard laser printer meet the hazardous materials
requirement, as do prints on paper or card stock from any other type of printer whose
negative impact on health is no worse than a laser printer.

4.10 Fingerprint Prints

4.10.1 Print Types Requirements

The printer shall have the capability to print a set of individual livescans or previously
scanned, individual inked fingerprints, life-size and in their correct printblock locations,
onto a standard ten-print fingerprint card (e.g., fingerprint card type FD-258), or print
onto blank 8.0 x 8.0 inch card stock, or print onto blank 8.5 x 11.0 inch plain paper. In
the case of printing fingerprints onto blank card stock or blank paper, the printer shall
also print the printblock boundary lines and labeling that normally appears on a standard
ten-print card.

The printer shall have the capability to print a previously scanned ten-print card, in its
entirety and life-size, onto blank 8.0 x 8.0 inch card stock, or onto blank 8.5 x 11.0 inch
plain paper.

NOTE: Printer margins for edge adjacent printblocks when printing on 8.0 x 8.0 inch
card stock may not exceed 10% of the image width dimensions. For an image 1.5 inches
wide, this means a margin of 0.15 inches or less. In worst case, truncation of card edges
is acceptable. Any shrinkage resulting in image reduction is unacceptable.

53
 The printer shall have the capability to print a single fingerprint, magnified up to 5 times
beyond life-size, onto 8.5 x 11.0 inch plain paper.

4.10.2 Requirements Compliance

The prints of the FP are visually inspected for conformance with the requirements.

4.11 Auxiliary Print Data

When printing in ten-print card format onto ten-print card stock, blank card stock, or
plain paper, the printer shall also have the capability to print labels, bar chart, step tablet,
and finger condition codes, all on the same print with the fingerprints. Figure 4-4
illustrates the printing of this auxiliary information; the following sections, 4.11.1
through 4.11.4, give the detailed requirements.

4.11.1 Labels

4.11.1.1 Requirements

When printing fingerprints in ten-print card format, the printing process shall have the
capability to print a character string of scanner information within the left four finger
plain impression printblock, and a character string of printer information within the right
four finger plain impression printblock. Each character string shall be printed along the
top inside edge of the respective printblock, in a type font and size that is large enough
for human readability without the aid of a magnifier, and small enough so as not to
unduly impinge on fingerprint structure; i.e., height of upper case letter or numeral in the
range: 0.067 inches to 0.095 inches.

Scanner Label Step Tablet Bar Chart Printer Label

6. L 7. L. INDEX 8. L. MIDDLE 9. L. RING 10. L.

THUMB LITTLE
Biojax model 1 with Acme LS-75 #123456 Ajax P-101 #6521 with Biojax sw v2 1/15/04 14:31:05

XX

LEFT FOUR FINGERS TAKEN SIMULTANEOUSLY L. THUMB R. RIGHT FOUR FINGERS TAKEN SIMULTANEOUSLY
THUMB

Condition Code
(XX = amputated)

54
Figure 4-4. Auxiliary Information in Ten-Print Card Format Print (Example Text)

The scanner information string shall include the following:

1) scanner make, model number, and serial number, if available
2) scanner system make, model number, and serial number, if available.

The printer information string shall include the following:

1) printer make, model number, and serial number, if available
2) name and version of the fingerprint printing algorithm, if available
3) date and time of printing.
The scanner and printer character strings shall be printed without a background, border,
or any other type of added surround.

Background:

Information for the scanner string can typically be obtained from the [EBTS] Type-2
Record Field identified as "IMA 2.067 - Image Capture Equipment", which includes
scanner system make, model number, and serial number.

A printer is certified as a combination of a specific brand/model printer and fingerprint

printing algorithm; the latter may also have a name or version designation.

Character string printing: a solid background (e.g., white) to the character string is
unacceptable because it would unnecessarily obliterate some parts of fingerprints on
some images. Individual characters with no background that overprint the fingerprint,
would obliterate a much smaller proportion of the fingerprint and are acceptable.
Printing the character strings in an open space created by off-setting printblocks 6-10
from printblocks 11-14 is unacceptable because it changes the dimensions of the standard
ten-print card format, and it cannot adequately accommodate fingerprints that stray
across printblock boundaries.

4.11.1.2 Requirements Compliance

The prints of the ten-print card format component of the FP are visually inspected for
conformance with the label content, size, and location requirements, aided by a magnifier
and precision ruler, where necessary.

4.11.2 Bar Chart

4.11.2.1 Requirements

55
When printing fingerprints in ten-print card format, the printing process shall have the
capability to print a bar chart, consisting of equally-spaced horizontal bars and equally-
spaced vertical bars, whose spatial frequency equals the required printer resolution.

The Bar Chart shall be positioned at the top edge within the right thumb plain impression
printblock and shall have a maximum width of 0.8 inches and a maximum height of
0.125 inches. The Bar Chart shall contain at least 10 parallel bars in each direction,
vertical and horizontal, with a bar length of at least 0.0625 inches (not necessarily the
same number of bars, or same bar length, in the two directions).

An optional, uniform mid-grey level patch may be included between the horizontal and
vertical bar components.

The bar chart shall be printed without a border or any other type of added surround.

Background:
For a 500 ppi printer requirement the limiting frequency is 250 cycles per inch, which
implies that 250 black bars per inch are printed, where the 0.002 inch width of an
individual bar is equal to the width of the white space between two bars.

If a mid-gray patch between the vertical and horizontal bar patterns appears to have the
same overall gray-level on the print as the two bar patterns, then this may indicate that
the printer gamma/highlight/lowlight settings are optimum and/or that the printer toner
supply was adequate for printing.

4.11.2.2 Requirements Compliance

The prints of the ten-print card format component of the FP are visually inspected for
conformance with the bar chart content, size, and location requirements, aided by a
magnifier and precision ruler, where necessary.

4.11.3 Step Tablet

4.11.3.1 Requirements

When printing fingerprints in ten-print card format, the printing process shall have the
capability to print a step tablet, consisting of two adjacent horizontal bands, each band
having 16 gray-levels. One band should progressively darken from left to right and the
other band should progressively darken from right to left. The 16 digital input gray-
levels corresponding to one band shall be identically the same as for the other band, and
both bands should cover the total gray-level range. This step tablet shall be positioned at
the top edge within the left thumb plain impression printblock and shall have a total
width between 0.5 inches and 0.8 inches, and a total height between 0.0625 inches and
0.125 inches.

56
The step tablet shall be printed without a border or any other type of surround.

Background:
If the top band and bottom band appear to be balanced on the print, i.e., the same mid-
gray-level appears in the middle of both the top and bottom bands, then this may indicate
that the printer gamma/highlight/lowlight settings are optimum.

4.11.3.2 Requirements Compliance

The prints of the ten-print card format component of the FP are visually inspected for
conformance with the step tablet content, size, and location requirements, aided by a
magnifier and precision ruler, where necessary.

4.11.4 Finger Condition Codes

4.11.4.1 Requirements

When printing fingerprints in ten-print card format, the printing process shall have the
capability to notate the presence of an abnormal finger condition in the appropriate
printed fingerprint block, for those cases where the [EBTS] Type-2 Record Field
identified as "AMP" (amputated or bandaged) is available, and/or for those cases where
similar information is available from other sources, such as a state system (possibly with
other notation codes).

4.11.4.2 Requirements Compliance

The prints of the ten-print card format component of the FP which have abnormal finger
conditions, and for which finger condition codes are available, are inspected to verify the
required presence and correct locations of the condition codes.

4.12 Fingerprint Quality

4.12.1 Requirements

The printer shall produce sufficient print quality to allow usable viewing of life-size
fingerprint prints under magnification to support fingerprint comparisons, i.e.,
identification or non-identification decisions. The lifesize print shall maintain its
sharpness and detail rendition structure up to at least 4X visual magnification, to the
extent that ridges, and ridge joints, bifurcations, and terminations that exist in the digital
image input to the printer, can be substantially discerned by the human observer on the
output print, without breaking-up and without getting lost in the noise. In addition, the

57
printing process shall not create significant false detail; e.g., shall not create ridge-like
patterns which did not exist in the input digital image.

4.12.2 Test Procedure

The prints of the FP are visually inspected with the aid of a magnifier, looking at items
such as:
- completeness of ridge patterns, in comparison to FP
- contrast, brightness, dynamic range, in comparison to FP
- artifacts, including false fingerprint structure, in comparison to FP
- capability of a life-size print to withstand 4X (or greater) magnification, without
the magnified image appearing to break-up or to get lost in the noise.

The prints of the FP may be quantitatively assessed by applying an appropriate image

quality metric to scans of the prints.

4.12.3 Requirements Compliance

Compliance with the requirements is obtained if the visual inspection of the FP prints,
and any appropriate quantitative measurement that may be applied, results in the
determination that sufficient image quality exists to allow usable viewing of life-size-
printed fingerprints up to 4X visual magnification, to the extent that a substantial portion
of the ridges, and ridge joints, bifurcations, and terminations that are in the FP, can be
discerned on the print without breaking-up and without being lost in the noise, and no
significant amount of false detail is discernible.

58
 SECTION 5

MOBILE ID

A mobile identification (ID) device is a livescanner viewed in the context of a portable

biometric acquisition station, i.e. one that is not intended to be stationary and hardwired
to a much larger system used for comparing or matching biometric samples. Since
mobile devices may satisfy a variety of collection modalities with differing image size
and accuracy requirements, there is a set of subject acquisition profiles for fingerprint
images, labeled 10-60, shown in Table 5-1. More information on the profiles and best
practices associated with mobile ID systems can be found in [MobileID].

5.1 Requirements

Image Size/Impression Type:

For a given fingerprint acquisition profile (FAP),16 the minimum image dimensions and
full range of simultaneous number of fingers specified in Table 5-1 shall be met. The
device shall be able to collect flat impressions. Rolled acquisition is optional.

Image Quality:
All requirements stated in either the [PIVspec] or Section 2 apply, as shown in Table 5-1.
When the IQS specification is PIV, then all requirements in the [PIVspec] shall be met.
When the IQS specification is App F, then all requirements in the [EBTS] Appendix F
Section 2 shall be met.

Table 5-1. Mobile ID IQS Requirements

Fingerprint
Acquisition Minimum Image Dimensions IQS Specification Simultaneous
Profile (WxH in inches) Requirements # of Fingers
(FAP)
10 0.5 x 0.65 PIV 1
20 0.6 x 0.8 PIV 1
30 0.8 x 1.0 PIV 1
40 1.6 x 1.5 PIV 1-2
45 1.6 x 1.5 App F 1-2
50 3.2 x 2.0 App F 1-4
60 3.2 x 3.0 App F 1-4

16 Prior to January 2012, Mobile ID terminology of Fingerprint Acquisition Profile (FAP) was referred to
as Subject Acquisition Profile (SAP).

59
5.2 Background

FAPs 10, 20 and 30 are for single finger sensors with FAP 10 having the same minimum
image dimensions as the Federal Information Processing Standard (FIPS) 201. FAPs 40
and above support simultaneous capture which is faster, reduces sequence errors and
produces higher quality images. As detailed in [MobileID] an agency will select a FAP
based on their specific requirements.

5.3 Test Procedures

Test procedures from [PIVtest] and Section 2 of this document will be applied,
respectively. Plain impressions containing typical simultaneous collections from both the
left and right hands of 10 subjects will be assessed for fingerprint image quality as
described in the above references. Some of the tested images will contain the maximum
number of simultaneous fingers listed in Table 5-1.

5.4 Requirements Compliance

Image Quality:
All requirements stated in Section 2 or [PIVtest], respectively, shall be met.

Image Size/Impression Type:

The requirement is met if the system has minimum capture dimensions in both width and
height and can capture the required number of fingers (Table 5-1) simultaneously.

60
 SECTION 6

FAST-TRACK CERTIFICATION

In certain defined cases full IQS certification testing, i.e., testing of all IQS requirements,
is not required. In these cases Fast Track testing, i.e., testing only some of the IQS
requirements, is sufficient.

6.1 Requirements

Fast Track certification testing is sufficient when:

• A vendor adds ‘value’ to an already-certified device, for example, by integrating

additional software which may include firmware (SW) and/or hardware which may
include firmware (HW), and repackaging the combination to create a value-added reseller
(VAR) label system. However, if there is a reasonable expectation that the added SW,
HW, or repackaging will affect the image quality performance of the original certified
device, then full certification testing would be required.

• A vendor makes relatively minor modifications to a previously certified device. For

example, a membrane is added to (or deleted from) a certified livescanner, an automatic
document feeder is added to a certified manual-feed cardscanner, or a 1000 ppi certified
scanner is operated at 500 ppi, using the same optics, sensor, and illumination.

Table 6-1 presents the test data requirements for some common Fast Track certification
scenarios; for test requirements for other scenarios contact the FBI.

In addition to the test data, the vendor seeking Fast Track certification must provide a
written statement to the FBI (letter or email) which affirms that the previously certified
fingerprint device has not been changed, with respect to device functions, hardware,
firmware, or software that could reasonably be expected to affect image quality
performance17. Specific to a scanner, the optics and optical layout, sensor, illumination,
image capture electronics and signal processing have not been changed and the maximum
capture area has not been increased.

No certification testing is necessary when:

• The original recipient of a certification wishes to change the model name and there are
no other changes to the certified product.

17Except for inherent image quality changes in specific situations, e.g., when recertifying a 1000
ppi scanner at 500 ppi.

61
• The original recipient of a certification wishes to repackage the device, if there is a
reasonable expectation that the repackaging will not affect the image quality performance
of the device. All device HW/SW components which may affect image quality
performance must remain the same as they were when originally certified. For example,
repackaging a device into a ruggedized cabinet, or repackaging a floor-standing device as
a desktop device by separating-out the host computer would not necessarily require
further testing, but changing the optical path or optical train of elements to accommodate
the repackaging would normally require retesting.

• A reseller of a certified device wishes to sell the device under its own label, or under the
original label. The certified device must remain intact, unmodified, and as a stand-alone
product with no added HW/SW. If relabeled by reseller, the certification is only valid
when that label does in fact contain the originally certified device, i.e., no blanket
certification for rebrands18.

• An end user receives a certified device to be used ‘as is’, without modification (an end-
user does not need its own certification).

Definition of Terms:

Vendor - generic term to include Original Equipment Manufacturer (OEM), reseller,

Value-Added Reseller (VAR), product assembler, systems integrator, and similar.

Full IQS Certification - a complete set of test data covering all IQS requirements is
submitted

Fast Track IQS Certification - a partial set of test data covering defined IQS requirements
is submitted

18If there is no Fast Track testing, the device will not be listed under the resellers name in the
FBI certification list. Instead it will remain listed / certified under the original vendor and device
name. A separate reseller listing requires some Fast Track data.

62
 Table 6-1. Fast Track Certification Procedures (Common Scenarios)

Fast Track Type Test Data to be Provided Requirements

Certification to FBI Compliance
Livescanner Vendor A incorporates Livescans from 5 subjects section 2.6
vendor B’s certified device (10 rolls & 4 plains, each
into vendor A’s value- subject).
added system.

Vendor adds (or deletes) Sinewave or bar target sections 2.1,

platen membrane to scans (target supplied by 2.3 & 2.6
certified device. vendor) and livescans from
5 subjects (10 rolls & 4
plains, each subject).

Cardscanner Vendor A incorporates Fifty 10-print card scans section 2.6

vendor B’s certified device (cards supplied by FBI)
into vendor A’s value-
added system.

Cardscanner vendor recertifies manual Fifty 10-print card scans section 2.6
with Automatic card scanner for use with (cards supplied by FBI)
Document ADF
Feeder (ADF)
Printer Vendor A incorporates Print of printer test target all subsections
vendor B’s certified device (target supplied by FBI) under section
into vendor A’s value- 4.0 pertaining
added system. to digital test
target
1000 ppi vendor recertifies its own Cardscanner: sections 2.1,
fingerprint fingerprint scanner in Sinewave target scans 2.3 & 2.6
scanner as 500 alternate operating mode (target supplied by vendor)
ppi fingerprint and ten 10-print card scans
scanner (cards supplied by FBI)
Livescanner:
Sinewave or bar target
scans (target supplied by
vendor) and livescans from
5 subjects (10 rolls & 4
plains, each subject)

63
6.2 Fast Track Permission

If the FBI agrees that the modifications present are not expected to adversely impact
image quality performance, then Fast Track testing can proceed.

In the case of certain optical modifications, it may be necessary to perform different

combinations of the IQS tests. For example, optics changes with no geometric effect
(addition of a membrane, or changed illumination) would not require new geometry tests.

Often Fast Track is applied during the testing of a single device that serves several
purposes. For example, a device that shares optics for 1000ppi and 500ppi should have
full testing for 1000ppi, and may use Fast Track testing (sine or bar-targets and reduced
number of livescans) for 500 ppi.

If the same optics are shared for tenprints and Identification Flats, then test targets for the
larger scan area must be provided. However if the sine/bar targets are small enough that
multiple placements are required, at least one should be placed in the area where the
rollscans are acquired.

6.3 Test Procedures

Perform each of the individual tests included in the Fast Track as described in sections 1-
5. When possible new test results will be compared to the previous certification test
results to ensure no unexplainable changes occur.

6.4 Requirements Compliance

The Fast Track requirement is complied with if the necessary paperwork is provided, Fast
Track permission is granted, the individual fast track tests meet compliance requirements,
and there is no unexplainable difference from the previous compliance testing data.

64
 SECTION 7

ADDITIONAL / OPTIONAL TESTS

In certain livescanner cases, additional data is requested.

7.1 Inked Card Comparison for New Designs

New sensing approaches that have not previously been certified or previously certified
sensing approaches that are implemented in a creative way may benefit from an inked
card comparison, where both inked cards and livescans from the new design are collected
from the same subjects for comparison. These images are inspected by fingerprint
examiners who provide commentary on ease of use and their comfort with the visual
appearance of the prints. If the images produced by the new design cause the examiner
discomfort that cannot be resolved or show signs of changes in print features, this
situation can be a significant deterrent in certifying the new design.

The test data consists of prints from the same people, taken with two (or three)
technologies:
1) ink on card;
2) new design; and
3) another previously certified livescan sensor (optionally).

The subject set should contain between five and ten people (50-100 prints), including a
distribution of gender, ethnicity, age, and, if possible, various skin conditions. There is
no percentage requirement for each type, just an attempt to get a broad distribution. One
way to include various skin conditions is to scan dry fingers, sweaty fingers (induced by
the wearing of a rubber glove), and fingers treated with an appropriate “ridge building”
substance to better prepare the skin (e.g., soaking in warm water, application of hand
lotion, hot sauce, or other commercially available products).19 In addition to various
stress conditions, include subject(s) with some easy-to-print fingers (e.g., wide ridges,
pores easily visible, uniform contrast across the print). Individual hands or fingers may
be prepared differently to obtain a wider range of skin conditions.

Personally identifying information (PII) should not be noted for any subject. Instead,
each subject should be assigned a number that is used consistently as a label across the
comparison collections. Documentation of the range and distribution of the above
characteristics (see section 7.3) is useful, but should only be provided to the extent
possible without revealing identifying information.

The vendor can collect their own inked cards (rolls and/or flats) after some
experimentation and practice. An inkpad and multiple 10-print cards are available from
the FBI for this purpose.

19 Any substance used to prepare the fingers must be removed before fingerprint capture.

65
7.2 Stress Imagery

During livescan testing, stress images are also requested, although their assessment does
not affect whether the device is recommended for certification. These images are used to
observe the ability of the device to collect impressions under simulated operational
conditions.

Submitted images fall into two distinct categories: those documenting the collection
conditions, which are color pictures taken with a digital camera, and those from the
livescanner. A stress image submission should include the following color and livescan
images (from the same finger of the same individual):

1. A livescan of the reference finger captured in normal operating conditions;

2. A color picture of the simulated sunlight setup, showing the display of a light
meter reading between 90,000 lux and 100,000 lux (see section 7.2.1 for
guidance);
3. A livescan when the device operates under simulated sunlight conditions (see
section 7.2.1 for guidance);
4. A color picture of the pad of the finger onto which an X was drawn with a black
marker (see section 7.2.2 for guidance); and
5. A livescan of the finger with a drawn black X, in normal operating conditions (see
section 7.2.2 for guidance).

7.2.1 Simulated Sunlight Collection Guidance

Stress imagery described in #2 and #3 here above refers to the same simulated sunlight
conditions. Guidance regarding the creation of these simulated sunlight conditions is as
follows:
• Use a light bulb whose emission spectrum is continuous, preferably close to that
of daylight, and encompasses the entire visible spectrum;
• Safely mount the light bulb a few centimeters above the platen, as illustrated in
Figure 7-1, possibly with the addition of a reflector. Illumination must be as
uniform as possible over both the platen of the livescanner and the light sensor;
• It is critical to adjust the distance between the light bulb/reflector and the platen of
the livescanner so that a basic light meter whose light sensor is aligned with the
height of the platen reads about 100,000 lux, but not less than 90,000 lux; and
• Care should be taken to avoid burns while inserting a finger between the light
bulb/reflector and the platen of the livescanner.

66
 Figure 7-1. Suggested Set-up to Simulate Full Sunlight Exposure.

7.2.2 Black X Collection Guidance

Stress imagery described in #4 and #5 here above refers to the same black ink X
conditions. Results are expected to vary based upon the skin reflectance of the subject
and the percentage of the finger pad covered by the black ink. Guidance regarding the
creation of these black ink X conditions is as follows:
• Choose a permanent black ink marker whose tip leaves a trace which is between 1
and 2.5 mm wide;
• Be sure to draw two line segments which intersect at their center, making an X
shape, in the center of the pad of the finger, so that the length of each segment is
about 3 times the width of the trace; and
• Obtain and send in the reference to the marker used for this experiment.

7.3 Metadata Documentation

Do not include PII, such as names, with any prints. However, because it is useful to
understand the distribution of certain properties that may impact fingerprint quality,
consider providing the following summary metadata information across the entire set of
fingerprint captures:

• How many subjects are of each gender?

• What is the age range covered and how many subjects are young versus old?
• How many subjects have a skin color other than white?
• How many subjects have good versus poor skin condition?

67
(blank page)

68
 LIST OF REFERENCES

ANSI/NIST – American National Standard for Information Systems - Data Format for
the Interchange of Fingerprint, Facial, & Other Biometric Information - Part 1,
ANSI/NIST-ITL 1-2011:Update 2015, Special Publication (NIST SP) - 500-290e3,
December 2015, National Institute of Standards and Technology (NIST), Gaithersburg,
MD. Latest release available at http://www.nist.gov/itl/iad/ig/ansi_standard.cfm

EBTS – Electronic Biometric Transmission Specification (EBTS), document number

NGI-DOC-01078-10.0.6, July 2016, Federal Bureau of Investigation, CJIS Division,
1000 Custer Hollow Rd., Clarksburg, WV 26306. Latest updates available at
https://www.fbibiospecs.cjis.gov/EBTS/Approved

MobileID – Mobile ID Device Best Practice Recommendation Version 2.0, November

2015, National Institute of Standards and Technology (NIST) Special Publication 500-
280v2, available at http://nvlpubs.nist.gov/nistpubs/SpecialPublications/NIST.SP.500-
280v2.pdf

PIVspec – Personal Identity Verification (PIV) Image Quality Specifications for Single
Finger Capture Devices, FBI Biometric Specifications, 10 July 2006, available at
https://www.fbibiospecs.cjis.gov/Document/Get?fileName=pivspec.pdf

PIVtest – Test Procedures for Verifying Image Quality Requirements for Personal
Identity Verification (PIV) Single Finger Capture Devices, MITRE Technical Report,
MTR060170, December 2006.

69
(blank page)

70
 APPENDIX A

LIVESCAN TESTING NOTES

The specific design of a livescan device largely determines which targets can or cannot
be successfully used. For example, for livescan devices which utilize the optical imaging
principle of Frustrated Internal Reflection (FIR), also known as Total Internal Reflection
(TIR), the type of two-dimensional test targets that can be successfully imaged depends
on factors such as the angle of incidence of the illumination with respect to the plane of
the prism surface used for fingerprint capture, and with respect to the image capture angle
of the optics and sensor.

Due to a host of potentially different designs, one device might be able to image the
standard Ronchi bar, uniform gray, and continuous tone sine wave targets on reflective
paper or mylar substrates, while another device might require use of film transmission
targets, and another device might not be able to adequately image continuous tone
targets, in which case chrome-on-glass binary targets might be suitable. Regarding the
latter case, continuous gray shading cannot be readily produced on chrome targets (or at
least it is prohibitively expensive), so a black/white bar target would be substituted for
the continuous tone sine wave target (generating CTF instead of MTF). Individual
uniform reflectance chrome targets (“neutral density filters”) might be used for linearity
assessment or to replace the uniform gray reflection targets, or images of a blank platen
might be obtained, varying the exposure time in stepped increments to effectively obtain
a series of uniform gray shades.

The livescan vendor seeking certification generally needs to perform some experimentation
to arrive at a suitable set of targets and procedures. The Ground Rule is to adhere as
closely as possible to the target types and procedures given in Section 2 of this Test
Procedures document, and stay as close as possible to the device's operational fingerprint
imaging mode; changing/backing-off from these, only as is necessary. For example, it may
be more convenient during testing to use an external light source, but this does not exercise
the device’s internal illumination as it is used in practice (spectral content, internal
reflections, etc.), so external lighting should be avoided.

The vendor’s proposed set of targets and test procedures should be communicated to the
FBI (see point-of-contact in Section 1), for comment and possible recommended changes,
with the goal in mind of obtaining a mutually agreed upon set of targets and procedures
that comprehensively serve the purpose of the testing. If a set of procedures are agreed
upon for one livescan model and, later, another model with the same basic design comes
along (from same vendor), the same set of test procedures can most likely be applied.

71
Some Additional Points:

• In some cases it may be necessary to apply a wetting agent to the targets, i.e. an
index-of-refraction matching fluid, or “liquid gate”. If a wetted sine wave target is
used, it may be necessary to recalibrate the target’s sine wave modulations and gray
patch density values in its wet state. Liquid-gating is obviously a messy process that
is prone to increased variability of results, with the possibility of increased
nonuniformities and emulsion swelling which may distort the target.

• Platen Coating (Membrane): if the livescan device uses a membrane over the glass
prism, usual purpose being to obtain better finger ridge contact to the prism surface,
then testing must be performed with the membrane in place. Live scanner
certification is “with membrane”, “without membrane”, or “with and without
membrane” (the latter requires two test sets).

• Roll Tracking Mechanism: The geometric integrity of a roll tracking mechanism is

not normally tested directly; i.e., the Ronchi bar test is normally performed in a static,
single image capture mode. However, roll tracking that has significant problems will
likely produce anomalies or artifacts in the roll fingerprint scans that are required to
be submitted; these will be assessed for such problems.

72
 APPENDIX B

COMMERCIAL SOURCES FOR TARGETS

This Appendix lists manufacturers and/or resellers of some of the test targets that could
be used in IQS certification. Table B-1 and the accompanying list of target vendors is
NOT all-inclusive and it is not intended to endorse one product over a competitor's
product. There may be other viable target types and commercial sources not listed here.

Table B-1. Test Targets

Target Model Type Target Target Target Substrate

Frequency Dimensions Material
Range (cy/mm) (millimeters)
M13-60-1X sine wave 0.188 to 12 48 x 70 photo paper

M15-60 sine wave 0.25 to 20 48 x 70 photo paper

M13-60-1/2X sine wave 0.25 to 12 24 x 35 photo paper

M-6 sine wave 0.375 to 80 46 x 70 film

M-7 sine wave 0.75 to 128 22 x 30 film

TC-12-1 bar 1 to 10 38 x 38 black chrome

on opel

____ Ronchi 1.0 cy/mm 100 x 100 reflective mylar,

typical chrome-on-glass

Munsell uniform ____ 216 x 279 coated paper

N9, N3 gray typical

IT8 step tablet ____ 156 x 104 photo paper, film

typical

Note: Due to the lack of 7 and 9 cy/mm bar patterns, the T-90 bar target is no longer
acceptable. The TC-12-1 is a possible alternative.

73
Commercial Target Sources:

Sine wave Applied Image, Inc.

Bar 1653 East Main Street
Ronchi Rochester, NY 14609
SIQT Telephone: (585) 482-0300
Step Tablet http://www.appliedimage.com
Precision Scale http://www.sinepatterns.com
(Includes Sine Patterns, LLC as
manufacturer of sine wave targets.)

Bar Precision Optical Imaging, Inc.

Ronchi 62 Honeoye Falls 5 Pts Road
SIQT Rush, NY 14543
Step Tablet Telephone: (585) 533-9133
http://www.precisionopticalimaging.com

Bar Edmund Scientific Corp.

Sine wave 101 East Gloucester Pike
Ronchi Barrington, NJ 08007
Telephone: (800) 363-1992
http://www.edmundoptics.com

Bar JML Optical Industries, Inc.

Sine wave 820 Linden Avenue
Ronchi Rochester, New York 14625-2710
Telephone: (585) 248-8900
http://www.jmloptical.com

Munsell Gray (N9, N3) X-Rite, Inc.

Step Tablet (Munsell Color Lab)
4300 44th Street SE
Grand Rapids, MI 49512
Telephone: (800) 622-2384
http://www.xrite.com
8.5” x 11” matte surface:
N9 - cat.#22408
N3 - cat.#22384

Bar Image Engineering

Ronchi Augustinusstrasse 9d
Step Tablet 50226 Frechen, Germany
Telephone: 49 (0) 22 34 - 91 21 41
http://www.image-engineering.de

74
Precision Scale Gurley Precision Instruments, Inc.
Bar 514 Fulton Street
Troy, NY 12181
Telephone: (800) 759-1844
http://www.gpi-optics.com/opto.htm

75
(blank page)

76
 APPENDIX C

GEOMETRIC ACCURACY MEASUREMENT

Ronchi Target Accuracy:

The geometric accuracy requirements, test methodology, and compliance verification for
a scanner are consistent with use of a 1.0 cy/mm Ronchi bar target that has less than a
0.25 percent error in bar center-to-bar center spacing. This error tolerance is readily
available from chrome-on-glass Ronchi targets, and can also be obtained from Ronchi
targets on a flexible mylar substrate. With this target error tolerance, it is not necessary
to incorporate target calibration data into the IQS compliance measurement
computations.

Even if the Ronchi target error is slightly larger than 0.25 percent, use of that target
without calibration data is acceptable as long as the scanner still meets the geometric
accuracy requirements. The reasoning is that it is highly improbable that the target errors
would be exactly in-phase with the scanner geometric errors, and of opposite sign, both
of which would be required for scanner errors to cancel target errors. If the geometric
accuracy requirement is still met with this target, it simply indicates that the combination
of target errors plus scanner errors are relatively small.

A printer uses a digital Ronchi bar target which has zero error when one period on the
target is exactly equal to a whole number of pixels. When printed at 500 ppi, a digital
Ronchi bar target with a period of exactly 18 pixels corresponds to a frequency on the
print of 1.0936 cy/mm, or a period of 0.0360 inches.

The following geometric accuracy assessment description is principally for a scanner;

where printer assessment differs, it is so noted. This assessment procedure for scanners
and printers is implemented in the geo program, which is on the IQS Test Tools CD.

Measurement Definitions and Approach:

1) The target is scanned such that the bars are aligned parallel to one of the two axes of
the scanner's detector array (detector rows or columns), to within 0.5 degrees, which then
defines the direction, vertical or horizontal.

2) Common terminology used in this Appendix is given in the following and is illustrated in
Figure C-1.

a) A measurement strip is a single, continuous strip in the target image, across

multiple parallel Ronchi bars, within which scanner resolution and geometric

77
 accuracy measurements are taken; e.g., 10 measurement strips are depicted in the
horizontal bar Ronchi target “A” in Figure 2-2.

b) A bar is an individual black stripe of the Ronchi target.

c) A bar segment is the portion of a bar that is contained in a measurement strip.

d) An edge line of a bar segment is the line at the boundary of the white background
and the black bar and is calculated from image data for that bar segment. There
are two edge lines for each bar segment.

d) The center line of a bar segment is the line running down the center of a bar,
parallel to the two edge lines for that bar segment.

e) The center point of a bar segment is the mid-point of the bar segment’s center
line.

f) In the following discussion references to row, top, and bottom pertain to horizontal
bars, with measurements in the vertical direction. The same test methodology is
applied to vertical bars, with measurements in the horizontal direction, by
substituting the terms column, left, and right, respectively.

Measurement strip
Center point
Edge line

Center line
Bar segment
Edge line
Bar segments in
Ronchi Bars measurement strip

Figure C-1. Illustration of Definitions Used For Ronchi Target Computations

78
3) The center line of a bar segment represents the line that is midway between the two
edge lines of that bar segment. All distances are measured between the center lines of
bars, because the more accurate tolerance for the scanner test Ronchi target is in terms of
bar center-to-bar center distance. Center lines are not calculated directly, however, they
are based on measurements of bar segment edge line locations. The steps for calculating
a bar segment’s edge lines are as follows:

a) The edge locations of each bar segment are calculated every fifth pixel along each
bar edge, for a distance of 0.25 inches for scanner measurements, yielding 26
points for each edge for a 500 ppi scanner and 52 points for each edge for a 1000
ppi scanner. For printer assessment (500 ppi print scanned at 1000 ppi) a
measurement strip distance of 0.20 inches is used, yielding 41 points for each
edge. Edge locations are determined using the super-resolution edge detection
method presented in Seitz20, Section 3: "Subpixel Resolution Edge Detection", to
locate edges of transition from black to white and white to black. This method
calculates the location of a point on an edge within 22-N of the true position, after
N iterations of a bisection method. In the geo program, each iteration is applied to
two different edge locations, x, and the process determines which of the two
locations yields the maximum value of the function r(x). The inferior edge
location, as indicated by a smaller r() value, is replaced by the midpoint (bisection)
of the two previous locations, and this process is then repeated with these two
locations. In order to overcome the possibility of dust or scratches altering the
expected edge locations, the initial edge range is set quite large, using x values
from -8 to 8. Because this range is so large, the initial search reduces the edge
range by only 1/8, over 32 iterations, which avoids missing the actual edge
location in this broad expanse. Once the range has been reduced in this manner,
then the normal Seitz bisection proceeds. To achieve (and probably exceed) the
desired 0.05 percent accuracy, N is set equal to 16.
The value of fi is the difference between consecutive image pixel values Ii and
Ii-1.

f i = I i - I i-1 (C-1)

4
p
r (x ) = å f i e -a ( x -i )2
2
,a = (C-2)
i = -4 4

b) Once the point positions along an edge of a bar segment have been determined, the
best fit line is calculated using these points in a linear, least squares regression to
determine the edge line. This line fitting is done separately for each edge of each
bar segment. The application of this line fitting requires that the y values

20Optical Superresolution using Solid-State Cameras and Digital Signal Processing, P. Seitz, July, 1988,
Optical Engineering, Vol. 27, No.7, pp. 535-540.

79
 correspond to the row positions for the horizontal Ronchi bars, and the y values
correspond to the column positions for the vertical Ronchi bars. Figure C-2 shows
an example of the 26 points along each edge of a bar, within a measurement strip,
that are used to determine the center line or center point of a bar segment scanned
at 500 ppi. The combination of the subpixel resolution method and linear, least
squares regression allows the position of the bar segment edge line to be accurately
determined to within a small fraction of a pixel, and this combination is robust in
the presence of dust and scratches on the target.

linear least squares regression (y = mx + b) run on each bar edge line:

bar edge line

y
Seitz
bar edge line

x
Figure C-2. Example of Calculating the Edge Lines of a Bar Segment

4) The distance between the centers of two bar segments in the same measurement strip
is effectively calculated as the perpendicular distance between the center point of one bar
segment and the center line of the other bar segment. As defined in equation C-3 and
illustrated in Figure C-3, the distance is actually computed from the bar edge slopes (m)
and intercepts (b) of the bar edges, and the center points (xc) of the bars.

D = cos(a )(y 0 - y1 ) (C-3)

where,
arctan (m10 ) + arctan(m11 )
a=
2
y0 =
( 00 01 ) c + ( b00 + b01 )
m +m x
2
(m +m ) x + ( b10 + b11 )
y1 = 10 11 c
2

xc is the middle row (column) of the measurement strip

80
 y
y= m00 x + b00

y= m01 x + b01

(Xc, Y0) y= m10 x + b10

a y= m11 x + b11

D = cos(a)|y0-y1|
where,
(Xc, Y1) a = [arctan(m10) + arctan(m11)] / 2
y0 = [(m00+m01)xc + b00 + b01] / 2
y1 = [(m10+m11)xc + b10 + b11] / 2
xc = 63rd column from leftside of
measurement strip
a

Figure C-3. The Distance Between a Point and a Line

The center point of a bar's centerline is located X columns in from the left edge of the bar
segment, where X = 63 for 500 ppi scans and X = 126 for 1000 ppi scans (for scanner
assessment). Figure C-4 illustrates the distance measurement between the centers of two
bar segments, for a target scanned at 500 ppi.

63 columns

bar center line

center-to-center distance

bar center line

Figure C-4. The Distance Between Two Bar Centers is the “One Bar Distance
Measurement”

81
5) As illustrated in Figure C-5, the first bar segment used in all measurements of a given
measurement strip is the topmost full-width bar in that strip, and the last bar segment
used is the bottommost full-width bar in that strip.

First full-width
bar segment

Measurement strip

Last full-width
bar segment

Figure C-5. Examples of First and Last Full-Width Bar Segments in a

Measurement Strip

Measurements Performed:

For each measurement strip, the scanner resolution is first measured, both to verify
compliance with the pixels per inch requirement and to establish a pixels per inch value
that can be used to convert subsequent geometric accuracy measurements to inches.
Next, measurements are made to test the across-bar geometric accuracy over a short
distance of one Ronchi bar cycle; then, measurements are made to test the across-bar
geometric accuracy over a longer distance of 6 Ronchi bar cycles.

The calculations necessary for verification of compliance with the resolution and
geometric accuracy requirements can be performed by the geo software, which is on the
IQS Test Tools CD. The input data file needed to run geo can be created by the
creategeofile software. The output of geo includes the across-bar results for each
individual 1-bar and 6-bar sample measurement (distance error and location), the
computed ppi for each 6-bar sample measurement, the along-bar distortion test results,
and summary results for all 1-bar and 6-bar measurements in each measurement direction
(vertical and/or horizontal). The viewgeo software accepts geo’s output file to create a
type of error map image, which is a convenient way to visualize the locations and
magnitudes of the geometric errors; e.g., highlight periodicities in the errors.

82
Resolution Measurements

The first set of measurements establishes the scanner resolution in the row and column
directions. Individual scanner resolution measurements are taken over every six bar
distance. Scanner resolution in pixels per inch (ppi) is calculated as the number of pixels
between the specified bar centers in a given measurement strip, multiplied by a
conversion factor, as given in equation C-4. The average resolution is then calculated for
each measurement strip.

ppi = K x (width of 6 cycles in pixels) (C-4)

where,
mm/inch 25.4
K= = = 4.233
number of cycles 6

Specific to printer assessment: printer target version A8 contains Ronchi bars that are
0.260 inches long when printed at 500 ppi. When a scanned print is input to the geo
software, a 0.200 inch section of this 0.260 inch length is used for each measurement
strip. Therefore, there is only one measurement strip in the 7 inch vertical bars Ronchi
segment and one in the 4.75 inch horizontal bars Ronchi segment. The average ppi is the
average taken over the entire strip length and this average ppi is then used in each
individual 6-bar distance error assessment within that strip.

Across-Bars Geometric Accuracy Measurements

The first set of measurements for geometric accuracy tests the short distance accuracy,
i.e., less than 0.07 inch distance. For each measurement strip, distance measurements are
made between the centers of each pair of adjacent bars, starting with the topmost
complete bar segment. Each distance measurement is converted from pixels to inches by
dividing the distance measurement (in pixels) by the previously computed average
resolution for the given directional measurement strip.

The second set of measurements tests the geometric accuracy for distances in the 0.07 to
1.5 inch range, by measuring the distance between six bars (6 bar cycles), which is a
target distance of 0.23622 inches for scanner assessment using 1.0 cy/mm Ronchi target
bars, or 0.21600 inches for printer assessment using 1.0936 cy/mm Ronchi target bars.
For every sixth bar in each measurement strip, distance measurements are made between
the centers of bar pairs having five bars between them, starting with the topmost
complete bar segment. Each distance measurement is converted from pixels to inches by
dividing the distance measurement (in pixels) by the previously computed average
resolution for the given directional measurement strip.

83
Along-Bar Geometric Accuracy Measurement (Distortion Measurement)

The bar center locations established for across-bar geometric accuracy assessment form
the basis of assessments for along-bar geometric accuracy assessments, i.e., distortion
assessments. As illustrated in Figure C-6, since the bar center locations at a, b, c, d, e, f
are known, from the previously computed across-bar measurements, then it only remains
to determine the largest difference in center locations over a 1.5 inch bar length and
compare it to the allowable maximum. Some additional measurement aspects:

- All comparable along-bar measurements are made within the image area corresponding
to a single, continuous Ronchi target.

- The along-bar measurements are applied to every bar.

- The difference in fractional rows between two bar centers is converted to inches by
dividing by the average ppi of the two corresponding measurement strips.

84
 1.5″
0.25″

a b c d e f

meas. avg. bar center

strip ppi location at row
1 501.3 a 10.0
2 500.2 b 11.3
3 498.5 c 14.2
4 500.8 d 16.2
5 502.2 e 15.3
6 504.1 f 14.5

largest vertical difference over 1.5ʺ horizontal distance:

[(rowd @ 16.2) - (rowa @ 10.0)] / 501.05 = 0.01237ʺ
0.01237ʺ meets spec, since it is less than 0.016ʺ

Figure C-6. Example of Along-Bar Distortion Measurement

85
(blank page)

86
 APPENDIX D

CONSTRUCTION OF STRATIFIED TEST CARD SET

This Appendix describes the procedure used to construct a stratified master card set21
from the FBI’s Fingerprint Card Master File (FCMF). The same procedure could be used
on other card sets that contain a range of card types, i.e., light, medium, and dark-inked
cards22.

Scanners are evaluated with a standard test set of fingerprint cards representing the range
of quality found in the master file. However, only a very low percentage of the FCMF
cards exhibit fewer than 128 gray-levels, so a straightforward sample test would require a
relatively large number of test cards to adequately test the scanner’s capability to handle
such low contrast cards. A stratified sampling protocol, although more complex,
overcomes this problem; i.e., stratified sampling avoids the necessity for a very large
number of test cards. A stratified sample provides much better discrimination by focusing
sample resources in the areas of interest.

Stratification of a 100 card set from the FCMF is based on the gray scale range23 of
finger three (right middle finger); strata definitions and sample sizes are given in Table
D-1.

Table D-1. Strata, Gray Scale Ranges, Sample Sizes for FCMF 100 Card Test Set

Stratum Finger 3 Number of Cards

Gray Scale Range
1 1 to 127 40
2 128 to 199 40
3 200 to 256 20

21Reference for generating stratified sample set: Survey Sampling, L. Kish, 1965, Wiley, NY, chapter 3:
“Stratified Sampling”, pp. 75-112.

22 In July 2015, a new stratified 50 card test set was constructed as the old card sets were retired. Selection
of the new 50 card set from a 480 card supply of deceased cards was performed by mimicking the retired
set's gray-level range and contrast distributions. The gray-level range was measured as the 90th - 10th
percentile gray-level difference. Both gray-level range and contrast metrics were calculated using the MTF
grayfinger software code.

23 In this Appendix it is the number of gray-levels that is being dealt with, which ranges from 1 to 256;
terminology used in other sections of this document refers to a pixel’s gray-level value, which ranges from
0 to 255.

87
The distribution of gray scale ranges for all 10 rolled fingerprints24 for the cards in each
of the three stratums is shown in Table D-2. For example, for the 40 cards in stratum 1
(400 rollprints) the vast majority of the individual finger images have less than 128 gray-
levels, which is the same range exhibited by all 40 finger 3 images. However, 75 of the
(non-finger 3) finger images from these 40 cards do contain between 128 and 199 gray-
levels. Note that these gray-level distributions are the result of card scan data gathered
from a scanner set-up with linear response and no adaptive gray-level processing. When
some form of adaptive gray-level processing is applied, one would expect an increase in
number of gray-levels in each stratum; e.g., many of the 325 samples in stratum 1, gray
range 1 to 127, would shift to the two higher gray ranges.

The weight, Wh, associated with stratum h, is the proportion of FCMF cards that fall
within that stratum. These weights were derived from gray scale measurements made on
4,685 cards randomly selected from the total FCMF population of approximately 30
million cards. The Wh weights for the three stratums (h=1,2,3) are given by:

W1 = (number of cards for which finger 3 has 1 to 127 gray-levels) ÷ 4685

W2 = (number of cards for which finger 3 has 128 to 199 gray-levels) ÷ 4685
W3 = (number of cards for which finger 3 has 200 to 256 gray-levels) ÷ 4685

Table D-2. Gray Scale Ranges in 100 Card FCMF Test Set

Stratum (h) Range: 1 to 127 Range: 128 to 199 Range: 200 to 256 Weight
(Wh)
1 325 75 0 0.0314
2 3 385 12 0.9628
3 0 64 136 0.0058

If the scanner under test is to be primarily used to scan cards from another master file of
fingerprint cards, and if it is known or assumed that this other master file has gray scale
characteristics significantly different from the FBI's FCMF, then a stratified random
sample set of test cards can be assembled from this other master file. To accomplish this,
a table analogous to Table D-2 is constructed, resulting in Table D-3. In Table D-3, the
parameter value n1,c is the number of subimages in the first stratum that exhibit a dynamic
range of 1 to 127 gray-levels. The parameter value n3,u is the number of subimages in the

24 The proportions of the various gray scale ranges in the four plain impression fingerprint blocks, which
were not used in constructing Table D-2, typically correspond to the proportions in the ten rolled
fingerprint blocks.

88
third stratum that exhibit a dynamic range of 200 to 256 gray-levels, and similarly for the
other cases.

Table D-3. Gray Scale Ranges for Card Test Set

Stratum Range c: Range s: Range u: Weight

(h) 1 to 127 128 to 199 200 to 256 (Wh)
Gray-Levels Gray-Levels Gray-Levels
1 n1,c n1,s n1,u W1

2 n2,c n2,s n2,u W2

3 n3,c n3,s n3,u W3

The proportion of subimages that have less than 128 gray-levels is equal to the weighted
average of within strata estimates; i.e., each stratum represents its share of the population.
This weighted average, Ac, is given by equation D-1:

where nh is the total number of subimages in stratum h

The weighted average for the proportion of subimages which have 200 to 256 gray-
levels, denoted Au, is similarly defined in equation D-2:

The specification requires that at least 99% of the strata-weighted images have 128 or
more gray-levels, i.e.;
(1 - Ac) ≥ 0.99

and at least 80% of the strata-weighted images have 200 or more gray-levels, i.e.;

Au > 0.80

where Ac and Au are the strata-weighted averages for the scanner output subimages.

89
(blank page)

90
 GLOSSARY

ADF Automatic Document Feeder

AFIS Automated Fingerprint Identification System
AIIM Association for Information and Image Management
AMP amputated or bandaged
ANSI American National Standards Institute
avg average
bpp bits per pixel
CD Compact Disk
CJIS Criminal Justice Information Services
CTF Contrast Transfer Function
cm centimeter
cy/mm cycles per millimeter
DFT Discrete Fourier Transform
EBTS Electronic Biometrics Transmission Specification
FAP Fingerprint Acquisition Profile
FBI Federal Bureau of Investigation
FCMF Fingerprint Card Master File
FICO Fingerprint Image Capture Operations
FIPS Federal Information Processing Standard
FIR Frustrated (Total) Internal Reflection
FP Fingerprint
H Horizontal
HW Hardware
IAFIS Integrated Automated Fingerprint Identification System
ID Identification
Inc Incorporated
IQS Image Quality Specification
ISO International Organization for Standardization
JPEG Joint Photographic Experts Group
LLC Limited Liability Corporation
mm millimeter
MTF Modulation Transfer Function
NIST National Institute of Standards and Technology
OEM Original Equipment Manufacturer
PGM Portable GrayMap
PII Personally Identifiable Information
PIV Personal Identity Verification
ppi pixels per inch
SAP Subject Acquisition Profile
SIQT Scanner Image Quality Test (target)
SNR Signal-to-Noise Ratio

91
spec specification
SW Software
TGT digital target for printer testing
TIFF Tagged Image File Format
V Vertical
VAR Value-Added Reseller
WSQ Wavelet Scalar Quantization

92