Side-Match Finite-State Vector Quantization with Adaptive Block Classification for Image Compression
Shinfeng D. Lin, Shih‐Chieh Shie · 2000
Introduction Image compres7[x has become esme tial for multimedia applications both ins[H67H and networktransUFH sra in recent years Vector quantization (VQ)has emergedas an e#cient and popular method inlos7jM[ data compres[xF smpres [1]--[5]. One feature of VQ is that high compresxFS ratios are pos2MM6 with relativelysy [j blocksHFHS unlike other coding techniques sn h as trans[xF coding. The other attractivenes of VQas sUU7 codings heme derives fromits optimality and these[U'2U[ y of hardware implementation of decoder. In VQsHH6HH the image to be codedis firs partitioned into non-overlapping blocks Each image blockis individually mapped to theclos26 codeword in the codebookbasj upon the minimumdisjHM[xH rule. Thes codewords areus[Hj' generated from a trainingsa byus27 the iterativeclus2[xHH algorithm sl h as the generalized LloydclusjH[xH algorithm [1]. CompresFM2 is achieved by replacingthes codewords with thecorres ondingindexes intransF[x76' ors[F' age applications Recon