Face Scanning
The goal of this project is to build high-quality statistical models for human faces. The applications of such models include face recognition, digital face aging, user interfaces, and face synthesis. We have built a scanning system that is able to capture images of human faces from different viewpoints under varying illumination. We have acquired face data from a large cross-section of the population, including scans of the same individuals multiple times a day and over a long period of time. Using this data we are building statistical models that capture variations in 3D face geometry, expression, illumination, and small-scale mesostructure (e.g. wrinkles, bumps, pores, etc.)
Background & Objective: Our research aims at building statistical models that capture how human faces vary between individuals and how they change over a period of time (e.g., within a day or a person's lifetime). There are several applications for such models, including: more robust face recognition, aging images of missing children/fugitives from a single photograph, digitally aging actors for entertainment purposes and re-animating faces of actors to match alternative sound tracks (e.g. foreign language dubs).
Technical Discussion: Our face scanner consists of a geodesic dome with 16 high-resolution color cameras and 150 computer-controlled white LED lights. The system captures the 3D geometry of the face using a commercial face scanning system. The data is used to compute a normal map and to estimate the diffuse reflectance at each surface point. We subtract the diffuse reflectance from the measured data and fit BRDF model to the remaining surface reflectance. We also measure the subsurface scattering of skin at few locations in the face using a special contact device and estimate skin translucency. We analyze the data for a large population of people of different gender, race, and age and under different external conditions (sweaty, cold, makeup, etc.).
Outside Collaborations: Tim Weyrich and Markus Gross, ETH Zurich; Jinho Lee and Raghu Machiraju, Ohio State University; Jovan Popovich and Daniel Vlasic, MIT; Henrik Wann Jensen, UC San Diego.
Contact: Jay Thornton
| Technical Reports: | |
| A Statistical Model for Synthesis of Detailed Facial Geometry | |
| Analysis of Human Faces using a Measurement-Based Skin Reflectance Model | |
| Measuring Skin Reflectance and Subsurface Scattering | |
| Estimation of 3D Faces and Illumination from Single Photographs Using a Bilinear Illumination Model | |
| Finding Optimal Views for 3D Face Shape Modeling | |
| Model-Based 3D Face Capture with Shape-from-Silhouettes | |
| Silhouette-Based 3D Face Shape Recovery | |
Technology Areas:
Imaging
Computer Vision
Modification Date: October 15, 2007
