Motion-compensated Spatio-temporal Context-based Arithmetic Coding for Full Scalable Video Compression

This paper is dedicated to the design of a fully scalable motioncompensated spatio-temporal subband video coding scheme. The main problem addressed is how to allow for a scalable capture and representation of the motion information and, at the same time, efficiently exploit the temporal correlation in the sequence. Two motion-compensated spatio-temporal decomposition structures are analyzed. Their respective amenability and limitations to provide an efficient and fully scalable representation of the information (both motion and texture) are investigated. This analysis led us to retain in the sequel a spatio-temporal signal decomposition approach proceeding first with a spatial decomposition. Two approaches for exploiting temporal dependencies in the different spatial bands are then studied. The first approach relies on classical motion-compensated temporal filtering, while the second uses motion-compensated spatio-temporal context-based arithmetic coding.