Bandwidth Compression of Image Signals

Abstract

The techniques of bandwidth compression of picture signals, in order to represent it by as few a binary number of bits as possible, have been extensively studied and reported in literature. This review describes some of these techniques, particularly in the context of encoding monochromatic images at bit rates less than 2 bits/Pel. The techniques have been broadly classified into Transform Coding, Predictive coding and Hybrid coding techniques. Transform coding basically converts blocks of raster scanned, sampled, two-dimensional set of picture elements into another set of transformed co-efficients. Non-adaptive and Adaptive KL, Discrete Foureir, Discrete cosine, Hadamard, Slant, Haar Transforms etc. or combinations thereof have been used for the purpose. Essentially the Adaptive quantization and Sample selection in the transform domain has resulted in coding picture areas of high activity with more number of bits and low activity areas with a lesser number of bits. Fairly good picture coding systems usin...