0% found this document useful (0 votes)
80 views1 page

Realistic Human Face Rendering For "The Matrix Reloaded": George Borshukov and J.P.Lewis ESC Entertainment

This document describes techniques for rendering realistic human faces for the movie "The Matrix Reloaded". It discusses using high resolution 3D face scans as the base geometry, extracting surface detail maps through displacement and bump mapping. It also describes an image-based approach to derive skin's bidirectional reflectance distribution function from photographs of actors under controlled lighting. Additionally, it presents a computationally inexpensive method to simulate subsurface scattering in skin through storing diffuse illumination in lightmaps and approximating diffusion separately for different color channels. Examples are shown of the rendered faces compared to a photograph, demonstrating photorealism.

Uploaded by

mikesfbay
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
80 views1 page

Realistic Human Face Rendering For "The Matrix Reloaded": George Borshukov and J.P.Lewis ESC Entertainment

This document describes techniques for rendering realistic human faces for the movie "The Matrix Reloaded". It discusses using high resolution 3D face scans as the base geometry, extracting surface detail maps through displacement and bump mapping. It also describes an image-based approach to derive skin's bidirectional reflectance distribution function from photographs of actors under controlled lighting. Additionally, it presents a computationally inexpensive method to simulate subsurface scattering in skin through storing diffuse illumination in lightmaps and approximating diffusion separately for different color channels. Examples are shown of the rendered faces compared to a photograph, demonstrating photorealism.

Uploaded by

mikesfbay
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 1

Realistic Human Face Rendering for “The Matrix Reloaded”

George Borshukov and J.P.Lewis


ESC Entertainment

Introduction inexpensive and fairly easy to implement. The result of the diffuse
illumination reflecting off the face in the camera direction is stored in a
The ultimate challenge in photorealistic computer graphics is 2-d light map (see Fig. 4). We then approximately simulate light
rendering believable human faces. We are trained to study the diffusion in the image domain. To simulate the different mean free path
human face since birth, so our brains are intimately familiar with for different light colors we vary the diffusion parameters for each color
every nuance and detail of what human skin is supposed look like. channel. For animations the lightmap needs to be computed at every
The challenge of rendering human skin is further complicated by frame, so our technique computes an appropriate lightmap resolution
some technical issues such as the fact that skin is a highly detailed depending on the size of the head in frame. For objects like ears where
surface with noticeable features in the order of ~100 microns and light can pass directly through, we employed a more traditional ray
the fact that skin is translucent. On The Matrix Reloaded we had to tracing approach to achieve the desired translucency effect.
create completely photorealistic renderings for most of the
principal actors including Keanu Reeves, Lawrence Fishborne, and Results
Hugo Weaving.
The above components are combined with our Universal Capture, real
Facial Surface Detail world Lighting Reconstruction technologies, and a ray tracer such as
mental ray to produce the synthetic images in Fig. 5 and 6. For
The geometry used for our rendering was based on a 100-micron comparison Fig. 7 shows a photograph of Keanu Reeves (Neo). The
resolution scan of a plaster cast mold of the actors’ faces. Arius3d bottom image is a fully virtual frame from The Matrix Reloaded.
provided the scanning technology. These scans had extremely high
polygonal counts (10 million triangles; see Fig. 1). To use these
models in production and preserve the detail we deployed the
following technique. A low-res ~5K quad model was constructed
using Paraform software. The model was given a UV
parameterization and then used as a subdivision surface. The high
resolution detail was extracted using the lightmapping feature of
the mental ray renderer combined with custom shaders that
performed ray tracing from the low-res subdivision surface model
to the high-detailed 10M triangle raw scan; the distance difference
is stored in a displacement map. We applied the low frequency
component of this map as displacement; the high frequency
component was applied using bump mapping.

Image-based Derivation of Skin BRDF


Our skin BRDF was derived using an image-based approach. In
Summer 2000 as part of the early stages of Matrix Reloaded R&D
we had a setup, which consisted of 30 still cameras arranged
around the actor’s head. Actors were photographed illuminated
with a series of light sources from different directions (see Fig. 2).
The setup was carefully color calibrated and photogrammetry was
used to precisely reconstruct the camera positions and head
placement with respect to each camera for each image. The
collected image data from each camera was brought into a
common UV space through reprojection using a cyberscan model
of the actor. This convenient space (see Fig. 3) allowed us to
analyze the skin reflectance properties for many incident and
outgoing light directions. We derived parameters for an
approximate analytical BRDF that consisted of a Lambertian
diffuse component and a modified Phong-like specular component
with a Fresnel-like effect. (We would like to acknowledge
Matthew Landauer for his contributions to this section).

Subsurface Scattering of Skin


As production progressed it became increasingly clear that realistic
skin rendering couldn’t be achieved without subsurface scattering
simulation. There are a number of published methods for rendering
translucent materials however they are all fairly complex, require
large amounts of CPU power and produce somewhat disappointing
results. To address this we developed a technique for producing the
appearance of subsurface scattering in skin that is computationally

You might also like