Image sequence coding at 64 kbps using vector quantization and block matching

The authors propose algorithms for compressing monochrome and color image sequences depicting videotelephone/videoconference scenery down to 64 kb/s. This rate is of interest as it is one of the primary rates within the ISDN (integrated services digital network) network architecture. Emphasis is given to synchronous transmission strategies in which successive frames are updated at a fixed bit rate determined by the frame rate and the channel rate. The basic structure of the underlying video coder is that of a temporal vector-based DPCM (differential pulse code modulation) coupled with conditional replenishment. The algorithms feature vector quantization in the residual domain and motion-compensated (using block matching), as well as previous-frame, prediction. Simulation results, using a 15-s-long color sequence, are given. These results include photographs of encoded frames and signal/noise ratio plots. Encoded frames of good visual quality were obtained using motion-compensated prediction. It is shown that median filtering can be useful in improving the perceptual quality of encoded frames