Hierarchical image sequence coding with tree structured vector quantization

The paper presents two different approaches to image sequence coding which exploit the spatial frequency statistics as well as the spatial and temporal correlation present in the video signal. The first approach is the pyramidal decomposition of the Motion Compensated Frame Difference (MCFD) signal in the frequency domain and the subsequent coding by unbalanced Tree Structured Vector Quantizers (TSVQ) designed to match the statistics of the frequency bands. The type of TSVQ used in this study possesses the advantage of low computational complexity with coding performance comparable to full-search vector quantization. The second approach is similar except that the order of motion estimation/compensation and pyramidal decomposition are interchanged. Motion estimation on each frequency band and on only the low-pass frequency band of each level of the hierarchy are both considered and compared. In this approach the low computational complexity of the block encoder is enhanced by the fact that processing of frequency bands can be independently implemented in parallel. Both approaches make use of the BFOS algorithm for rate allocation to the frequency bands. Miss America image sequence has been coded at an average PSNR of 39.17dB and an average rate of O.35bpp with the first approach. Implementation of the second approach resulted in an average PSNR of 38.70dB at an average rate of O.28bpp. Both approaches are suitable for multi-rate video conference applications.