Very low bit rate video coding with predictive-transform schemes

The state of the art in digital video coding is represented by a class of standardized algorithms that have in common the basic techniques of redundancy reduction, exploiting temporal correlation with prediction and spatial correlation with transformation. The development of audio-visual signal coding for public switched telephone networks (PSTN), local area networks (LAN), mobile networks (UMTS), raised a strong interest in the implementation of image compression algorithms down to very low bit rates (VLBR), of the order of a few tens of kbit/s. Current research activity in VLBR image coding is going on in two directions: the “short term ” goal, that is the optimization of standard algorithms (namely, H.261 and MPEG-1) for the target formats, picture rates and bit rates; and the “long term” goal, i.e. the development of sophisticated methods for analysis-synthesis coding schemes, without constraints of compatibility with the first“ generation” algorithms. The article reports the results of VLBR coding of video for audio-visual communication applications, using traditional predictive-transform schemes