GIT-GVU-93-01

Exploiting Spatio-Temporal Coherence in Ray-Traced Animation Frames

Stephen J. Adelson
Larry F. Hodges

The majority of computer graphics images are generated as part of a sequence of animation frames. The usual approach for producing these images, however, is to render each frame in an animation as if it were a single, isolated image. We describe a technique that exploits spatial-temporal coherence between frames to speed up the generation of a ray-traced animation sequence. The concept is that the information gained when ray-tracing a particular frame in a sequence may be used to speed up the ray-tracing of other nearby frames. Based on a stereoscopic generation technique that generates the second half of a stereo pair for as little as 5% of the effort required to fully ray-trace the first half, this approach promises significant savings when ray-tracing animations with certain characteristics. Our algorithm is tested on an animation with many reflective surfaces, yet it still attains an overall 60% savings over full ray-tracing in terms of the number of rays solved.

Display algorithms, ray-traced animation, spatial coherence, temporal coherence

PDF

Postscript