Evolutionary clustering based vector quantization and SPIHT coding for image compression

Taking both the correlations among and within wavelet subbands of images into account, in this paper we proposed an evolutionary clustering based vector quantization (VQ) and set partitioning in hierarchical trees (SPIHT) coding method for image compression. One-step gradient descent genetic algorithm (OSGD-GA) is designed for optimizing the codebooks of the low-frequency wavelet coefficient by defining the importance degree of each coefficient and utilizing fuzzy membership to address the automatic clustering. This new VQ technology exploits the global searching capability of OSGD-GA and can automatically obtain contextual constraints on membership condition by weighted average method of the importance, so it can overcome the drawbacks of classical clustering algorithm. Then the scalar quantization followed by SPIHT coding algorithm is employed for the high-frequency wavelet coefficients. Some simulational experiments are taken to investigate the performance of the proposed method. The results show that our proposed method not only brings about some new ideas in combining the evolutionary clustering based VQ and SPIHT coding, but also yields an improvement of PSNR to the greatest 0.66 dB over SPIHT algorithm.