Wavelet-based video compression at low bit rates

Publisher Summary This chapter discusses a video coding method based on the use of "I" and "B" frames within a logarithmic decimation structure. Spatial decimation is performed by wavelet filtering and the quantized output is processed using an entropy coder similar to that used in JPEG. The key areas developed in this chapter are open-loop coding, windowed motion-vector compensation, temporal weighting, and noise coding. The chapter describes new and established techniques developed for low bit-rate coding of motion images. The conclusions of the simulation studies are that the conventional technique of predictive coding has severe quality limitations at these bit rates and that wavelet open-loop coding produces much better picture quality. It was shown that the spatial quantizer profile and temporal weighting add considerable coding benefit. In addition, soft window MVC results in a distinct benefit to the picture quality. An improved form of noise coding has been developed, which reduces isolated noise elements rather than picture information. One area not explored in the chapter is the coding delay that occurs as a result of the improved coding efficiency. The total loop delay of approximately 1 second may be difficult to accept in practice. This is an area for future study.