Automatic object segmentation algorithms for sprite coding using MPEG-4

Object-based video coding, as standardized in MPEG-4 Part 2, can result in superior performance in comparison to common hybrid motion-compensated DCT-based approaches. We consider sprite coding which increases significantly the objective as well as the subjective quality of the coded video. The main challenge of this approach is the pre-segmentation of the video and the video content itself. To apply sprite coding, the input video has to be firstly segmented into foreground and background objects. We evaluate automatic object segmentation methods based on global motion estimation and background sprite generation. These algorithms are evaluated using the standardized MPEG-4 visual main profile (sprite coding).