Facial Reflectance Field Demo
Accompanies the Paper

Acquiring the Reflectance Field of a Human Face by Paul Debevec, Tim Hawkins, Chris Tchou, Haarm-Pieter Duiker, Westley Sarokin, and Mark Sagar.

Demo software written by Chris Tchou and Dan Maas. Developed at the USC Institute for Creative Technologies.

This demo allows you to virtually relight real faces. You can light the subject as if they were in a variety of real lighting environments captured from around the world, or position and adjust your own virtual lights to achieve whatever effect you desire.

Different subjects are selected by clicking on the thumbnails at the left (additional face datasets may be downloaded individually; see below). Clicking on Environments at the upper right displays a choice of several different environments. The current lighting environment is displayed in a low-resolution panoramic image below the face. Clicking on the yellow arrows to the right or left of this image will make the environment spin continuously around the subject.

Clicking on Lights at the upper right displays a reference sphere and three virtual lights. These lights may be moved by dragging them around the sphere. When a light is selected (yellow), you may adjust its brightness, color, and diffusion using the controls at bottom right. The panoramic image below the face shows the cumulative lighting environment due to the three virtual lights.

The program computes a weighted sum of 512 images, each capturing the appearance of the face when illuminated from a particular direction. The relighting computation is performed in real-time directly on the discrete cosine transform coefficients of the data compressed in lighting-direction space. For more information, please see the corresponding research paper and website.


Download and unzip the demo into a directory, then run face.exe (Windows) or face (Linux). To add a new face dataset, simply unzip the package (a new subdirectory will be created under faces). We recommend that you not load more than one face dataset per 40MB of RAM in your system.

Download the Software


Linux x86


Download Additional Face Datasets

Light Stage 1.0 Datasets

Paul (19M)
H.P. (25M)
Tim (21M)
Eric (18M)
Westley (25M)
Jessica (14M)

Light Stage 2.0 Datasets

Andrew (18M)
Emily (21M)
George (70M)
Headdress (33M)

For more information on the "George" and "Headdress" datasets, see: Hawkins et al, " A Photometric Approach to Digitizing Cultural Artifacts", VAST 2001.

Send us comments, suggestions, and patches.

Footer With Address And Phones