Adaptive vector quantization of video signals with motion compensation and spatial masking

An efficient adaptive scheme is proposed for low-bit-rate video sequence coding. The redundancy within adjacent video frames is exploited by motion-compensated interframe prediction, and the motion-compensated frame difference (MCFD) signal is vector quantized adaptively. The MCFD and the motion information are transmitted to the receiver to reconstruct the frames. It is noted that the dependence of the vector codebooks on the set of training signals is much less for the MCFD signals than the original video frames. The MCFD signals are classified into the five groups according to the local variances. A multicodebook vector quantizer is designed that is based on all groups. Some simple masking properties are also used in the design of the vector quantizer. An average signal-to-noise ratio of 37.85 dB is obtained for their highly active CINDY video sequence with an average bit rate of 0.47 b/pixel over 40 frames