Fundamentals of Lossy Video Compression

Abstract

In the previous chapter, we introduced some of the basic concepts in the theory of lossy compression of still-images. These concepts can be readily applied to the compression of image sequences (such as video) as well, by simply treating each image in the sequence as a still-image. This approach is inherently simple; however, it does not provide significant compression. For instance, consider an uncompressed image sequence at a data rate of 166 Mbits per second. Using a DCT-based coding system on an image-by-image basis, for good image fidelity, one can achieve close to 12:1 data compression. Thus, the compressed bit rate is close to 14 Mbits per second, which is too high for most practical uses. For instance, the bandwidth for digital TV broadcasting is close to 4 to 6 Mbits per second. In this case, we need a compression ratio around 41:1. In another example, the typical output rate of a CD-ROM drive is close to 1.5 Mbits per second. In this case, we need a compression ratio around 110:1. Hence, in practical applications with image sequences, we require at least four times better compression ratios than those we can achieve with coders for still-images.