Realtime Synthesis of Super-Realistic Moving Goldfish Using Small-Scale Parallel Transputers

Abstract

Parallel computer is one of important key technologies for achieving visual realism and natural behavior in realtime. This paper describes a transputer-based small-scale parallel computing system for synthesizing texture mapped animated goldfishes in realtime. It employs the techniques of a deformable three-dimensional wireframe model, texture mapping and linear movement compensation to generate realistic figure and natural motion. General purpose microprocessors are effectively used to realize a practical parallel computing system especially for visual data. Each transputer is controlled by small-sized packets, and responsible for generating own goldfish image. Network links connecting these transputers work in parallel to this purpose. This paper shows the result of the empirical trace of parallel computation on our prototype system. The total texture mapping performance for four goldfish synthesis is faster than 10 frames/sec on four transputers. This prototype system can also change the color of moving goldfishes with natural texture continuously by a luminance mapping technique.