BAP sparsing: a novel approach to MPEG-4 body animation parameter compression

Abstract

The MPEG-4 standard includes support not only for natural video and audio, but also for synthetic graphics and sounds. In the MPEG-4 specifications, body animation parameters (BAPs) and body definition parameters (BDPs) allow virtual bodies and their animation to be compressed using a standard compression pipeline comprising of quantization, predictive encoding and arithmetic coding of these parameters. In this paper, we propose and implement a new stage within the standard prediction-based compression pipeline for the BAPs, termed as BAP sparsing. BAP sparsing compresses a complete block of motion data, consisting of an initial 1-frame followed by subsequent P-frames, required for creating the animation. It exploits the inherent hierarchical structure of the human skeletal model to intelligently drop and modify the P-frames, while preserving animation quality. BAP sparsing is shown to result in superior compression of the BAP data, with negligible loss in the motion animation quality. It is also shown to result in a lower network throughput requirement and fewer CPU cycles on the client end to create the animation. The proposed method is particularly well suited for animation using BAP data.