Transform vector quantization of images in one dimension

Abstract

In this paper there is enclosed a description of the hybrid system for black and white (256 by 256 pixels) images compression. The system includes the following procedures: image decomposition (8x8, 16x16 blocks), DCT transformation, “zig-zag” scanning, product code vector quantization (one-dimensional block) and a bit allocation. The standard LBG algorithm for codebook design has been enriched with the simulated annealing procedure for avoiding the local minima. Standard vector quantization in two-dimensional transform space has been investigated for comparison.