Core Topics in Geographic Information Science

Assignment

On

Section B: Image Analysis for Geoinformatics

Submitted by

Name: Roquia Salam

Student Id Number: 542552

Email address: rsalam@uni-muenster.de

Submitted to

Professor Benjamin Risse

MSc in Geospatial Technologies

Page 1 of 6
Question: What are the key differences between "classical / analytical image analysis",
"geometric image analysis" and "machine learning-based image analysis"? For which tasks
/ goals are these different paradigms used, what are their strengths and weaknesses and how
are they related to (well-known) GI applications?

1. Introduction

Image analysis (IA) refers to extracting valuable information from digital images, using digital
image processing techniques (Solomon and Breckon, 2011). In the present day, IA has gained
significant importance due to its capacity to perform rapid, non-invasive, and cost-effective
analysis of products and processes (Prats-Montalbán et al., 2011). There are three main distinct
paradigms in IA and computer vision (CV) which are: classical, geometric, and machine learning-
based image analysis. Each approach has its unique strengths and weaknesses, making them
suitable for different tasks and applications in Geoinformatics. While classical image analysis and
geometric analysis have been foundational in the field (Richards and Richards, 2022), machine
learning (ML) based image analysis has emerged as a dominant approach due to its ability to
handle complex data and automate various image analysis tasks (Zhang et al., 2022), transforming
the landscape of geospatial analysis and applications.

2. Key differences between classical, geometric, and ML-based IA

2.1 Methodology:
In classical image analysis, conventional image processing techniques and algorithms, such
as mathematical and statistical operations, are utilized to directly manipulate pixel values
and spatial relationships within the image (Tan et al., 2021). Whereas, geometric image
analysis emphasizes examining the geometric attributes of objects in the image,
encompassing factors like shape, size, orientation, and spatial relationships (Jorge, 2023).
While ML-based image analysis relies on algorithms that automatically learn patterns and
representations from labeled training data (Zhang et al., 2022). It uses statistical techniques
to generalize and make predictions on new, unseen data.
2.2 Automation:
Classical and geometric image analysis techniques frequently necessitate manual
adjustment and the design of processing steps, resulting in lower automation compared to

Page 2 of 6
 machine learning-based approaches that can automatically learn from data (Hegde et al.,
2019).

3. Goals for using these three paradigms

i. Classical image analysis finds extensive application in fundamental image processing
tasks, encompassing image filtering, edge detection, image enhancement, thresholding, and
morphological operations. Additionally, it is harnessed for image restoration and noise
reduction purposes (Fernandes et al., 2020).
ii. Geometric image analysis serves various applications, including object recognition, image
registration, geometric measurements, and computer-aided design (CAD) applications. It
proves to be of significant value in extracting geometric features and contours from images
(Fernandes et al., 2020).
iii. ML-based image analysis assumes a dominant role in various tasks, such as object
detection, image classification, semantic segmentation, image generation, and content-
based image retrieval. Furthermore, it is widely adopted in computer vision applications,
such as facial recognition and image captioning, showcasing its versatility and prominence
in contemporary image analysis endeavors (Dev et al., 2016).

4. Strengths of these three paradigms

i. The classical approach proves advantageous for tasks that involve well-defined image
properties and explicit processing steps. It exhibits computational efficiency and can
effectively handle small datasets (Sabins Jr et al.,2020).
ii. On the other hand, the geometric approach excels in comprehending the geometric
structure and arrangement of objects within an image, making it a suitable choice for tasks
that entail shape analysis and spatial relationships (Jorje, 2023).
iii. In contrast, ML approaches demonstrate proficiency in handling complex and high-
dimensional data, rendering them suitable for tasks characterized by substantial variations
and extensive datasets. Once trained, ML models exhibit high automation and can adapt
adeptly to new patterns and variations in the data (Mahmood et al., 2022).

Page 3 of 6
 5. Weakness of these three paradigms
i. The classical image analysis may exhibit limitations in effectively addressing intricate
patterns or variations. This approach often necessitates manual adjustments and the explicit
design of processing steps, leading to lower levels of automation when compared to
machine learning methodologies (O’Mahony et al., 2020).
ii. Geometric image analysis, on the other hand, may encounter challenges in tasks that
demand intricate feature extraction or recognition in the presence of substantial variability
(O’Mahony et al., 2020).
iii. ML-based image analysis, while powerful, demands considerable amounts of labeled
training data for optimal performance. The quality of outcomes heavily relies on the
excellence and representativeness of the training dataset. Furthermore, complex models
may necessitate significant computational resources to execute effectively (Zhang et al.,
2022a).

6. Relationship of these three paradigms to Geoinformatics (GI)

Applications

GI applications heavily rely on image analysis techniques to extract valuable information from
geospatial imagery. Three prominent approaches, namely classical/analytical image analysis,
geometric image analysis, and ML-based image analysis, play pivotal roles in advancing GI
applications. Some of the applications are noted below:

i. Classical image analysis plays a vital role in numerous GI applications, specifically in

fundamental image preprocessing and enhancement tasks. These tasks are of utmost
importance in various areas like remote sensing, land cover mapping, and image fusion for
Geographic Information System (GIS) data preparation (Prats-Montalbán et al., 2011).
ii. Geometric image analysis assumes a critical role in GI applications that involve object
recognition and registration. It facilitates the identification and alignment of geographical
features, such as buildings, roads, and coastlines, within satellite imagery. The spatial
understanding provided by geometric analysis aids in tasks like land cover mapping,
infrastructure assessment, and urban planning (Prats-Montalbán et al., 2011).

Page 4 of 6
 iii. ML-based image analysis has brought about a revolutionary impact on GI applications,
particularly in tasks like land cover classification, change detection, and object detection
from aerial and satellite imagery. This approach enables highly accurate and automated
analysis of extensive geospatial data, ushering in new possibilities for advanced data
processing and interpretation in the field of geographic information science (Zhang et al.,
2022a).

7. Conclusion

The three image analysis approaches discussed in this paper are not mutually exclusive but rather
complementary. Classical image analysis provides fundamental preprocessing and enhancement
capabilities, which are often necessary as a first step in GI applications. Geometric image analysis
augments the spatial understanding of geospatial data, while ML-based image analysis enables
automated and sophisticated analysis of vast and complex datasets. The integration of these
approaches can yield enhanced accuracy, efficiency, and automation in GI applications, thereby
advancing the field of geographic information science.

References
Dev, S., Wen, B., Lee, Y. H., & Winkler, S. (2016). Ground-based image analysis: A tutorial on
machine-learning techniques and applications. IEEE Geoscience and Remote Sensing
Magazine, 4(2), 79-93.

Fernandes, A. F. A., Dórea, J. R. R., & Rosa, G. J. D. M. (2020). Image analysis and computer
vision applications in animal sciences: an overview. Frontiers in Veterinary Science, 7,
551269.

Hegde, R. B., Prasad, K., Hebbar, H., & Singh, B. M. K. (2019). Comparison of traditional image
processing and deep learning approaches for classification of white blood cells in
peripheral blood smear images. Biocybernetics and Biomedical Engineering, 39(2), 382-
392.

Page 5 of 6
Jorje, L. (March, 2023). Geometric Transformation in Image Processing: Basics, Applications, and
CronJ as an Expert. [URL: https://medium.com/@livajorge7/geometric-transformation-in-
image-processing-basics-applications-and-cronj-as-an-expert-f06417193695 ]

Mehmood, M., Shahzad, A., Zafar, B., Shabbir, A., & Ali, N. (2022). Remote sensing image
classification: A comprehensive review and applications. Mathematical Problems in
Engineering, 2022, 1-24.

O’Mahony, N., Campbell, S., Carvalho, A., Harapanahalli, S., Hernandez, G. V., Krpalkova, L., ...
& Walsh, J. (2020). Deep learning vs. traditional computer vision. In Advances in
Computer Vision: Proceedings of the 2019 Computer Vision Conference (CVC), Volume
1 1 (pp. 128-144). Springer International Publishing.

Prats-Montalbán, J. M., de Juan, A., & Ferrer, A. (2011). Multivariate image analysis: A review
with applications. Chemometrics and intelligent laboratory systems, 107(1), 1-23.

Richards, J. A., & Richards, J. A. (2022). Remote sensing digital image analysis (Vol. 5).
Berlin/Heidelberg, Germany: springer.

Sabins Jr, F. F., & Ellis, J. M. (2020). Remote sensing: Principles, interpretation, and applications.
Waveland Press.

Solomon, C., & Breckon, T. (2011). Fundamentals of Digital Image Processing: A practical
approach with examples in Matlab. John Wiley & Sons.

Tan, C. K., Goh, C. M., Aluwee, S. A. Z. B. S., Khor, S. W., & Chai, M. T. (2021, September).
Malaria Parasite Detection using Residual Attention U-Net. In 2021 IEEE International
Conference on Signal and Image Processing Applications (ICSIPA) (pp. 122-127). IEEE.

Zhang, J., Li, C., Rahaman, M. M., Yao, Y., Ma, P., Zhang, J., ... & Grzegorzek, M. (2022). A
comprehensive review of image analysis methods for microorganism counting: from
classical image processing to deep learning approaches. Artificial Intelligence Review, 1-
70.

Zhang, Y., Liu, J., & Shen, W. (2022a). A review of ensemble learning algorithms used in remote
sensing applications. Applied Sciences, 12(17), 8654.

Page 6 of 6