Block truncation coding by using genetic algorithm

Vector quantization (VQ) is an important method of lossy image compression. The basic prerequisite of VQ is that the codebook must have representability to ensure the quality of the recovery image. In this paper, we propose a new codebook design approach using genetic algorithm. Our method will split a gray-level image into blocks of 4*4 size and simplify these blocks to enhance the representability of our codebook besides, we modify some GA operations to avoid illegal chromosomes. The experimental results show that our method has better representability and generalization than conventional method. As for the quality of recovery images, our method also outperforms the conventional method.