Scalable video coding using 3-D subband velocity coding and multirate quantization

The authors have developed a class of scalable video compression algorithms capable of producing one compressed full-resolution video bit stream with various subsets corresponding to different resolutions/rates of the same video sequence. The strategy is to use velocity-based 3-D subband coding to decompose the video into different spatiotemporal resolutions, and to multirate quantize each subband in order to achieve a large number of bit rates with fine granularity. This scheme results in a minimum rate of 55 kbit/s and maximum rate of 3.024 Mbit/s for the ´Football´ sequence with 33 intermediate rates. The MSE (mean square error) of the proposed system is 11-17% higher than that of MPEG1 (Motion Picture Expert Group 1) for the three bit rates tested.<>