A Codebook Compression Method for Vector Quantization Algorithm

Abstract

This study introduces a novel approach to enhance the compression ratio of the vector quantization (VQ) algorithm by specifically targeting the compression of its codebook. The VQ algorithm typically generates an index matrix and a codebook to represent compressed images. The proposed method focuses on reducing the size of the codebook, which comprises N codewords, each with elements quantized into four levels. Each 8-bit element in a codeword is compressed to 2-bits, and the encoded codeword is accompanied by the minimum value and a threshold value in the codebook. Experimental results on benchmark color images, such as baboon, airplane, Lena, and others, demonstrate a significant reduction of 62.50% in the size of the VQ codebook.