MPEG-4 standardized methods for the compression of arbitrarily shaped video objects

MPEG-4 is the most recent standard for audio-visual representation to be published by the International Organization for Standardization. One of the many new features of MPEG-4 is its ability to represent two-dimensional video objects of arbitrary shape. For this purpose, MPEG-4 uses the conventional motion-compensated discrete cosine transform syntax for color/texture coding and augments this with an explicit compressed representation of the video object's shape. This paper is intended as a tutorial in the means of encoding and decoding arbitrarily shaped video objects as specified by MPEG-4. The major emphasis of the paper is on explaining the compression technology associated with the normative shape representation, i.e., block-based context-based arithmetic encoding, but some new aspects associated with arbitrarily shaped texture coding are also highlighted. The MPEG-4 specifications are presented in an informal way, and the motivations underlying the algorithm are clarified. In addition, effective methods are suggested for performing many of the nonnormative encoding tasks, and several encoding performance tradeoffs are illustrated.