Rendering, or synthesizing images from models of 3D scenes, is a major field of study in computer graphics. In many contexts, the goal of rendering is to produce images that are visually or even quantitatively indistinguishable from photographs. Achieving this goal requires accuracy in all parts of the rendering system: the algorithm, the geometric model, and the model of light reflection. Accuracy is especially important when the viewer is familiar with the object being rendered and when complex natural materials are involved. One long-standing problem in computer graphics that exemplifies all of these difficulties is the realistic rendering of moving human faces.
Understanding reflectance is a requirement for rendering real materials faithfully, and the best way to achieve this understanding is to begin with measurements. This talk will present a technique that uses digital photographs to measure the reflectance (specifically, the BRDF) of materials for use in rendering. The advantages of the technique are speed, flexibility, easy implementation, and accuracy. Results that demonstrate the accuracy of the technique will be presented, as will measurements of materials that were previously very difficult to measure, including human skin. Finally, the use of these measurements in rendering realistic human faces will be discussed.
This work was done with various collaborators, including Steve Westin, Eric Lafortune, and Brian Guenter.