ENTROPY-CONSTRAINED PREDICTIVE RESIDUAL VECTOR QUANTIZATION

A major problem with a vector-quantization-based image compression sch eme is its codebook search complexity. Recently, a new vector quantiza tion (VQ) scheme called the predictive residual vector quantizer (PRVQ ) was proposed, which gives performance very close to that of the pred ictive vector quantizer (PVQ) with very low search complexity. This pa per presents a new variable-rate VQ scheme called the entropy-constrai ned PRVQ (EC-PRVQ), which is designed by imposing a constraint on the output entropy of the PRVQ. We emphasized the design of the EC-PRVQ fo r bit rates ranging from 0.2 to 1.00 bits per pixel. This corresponds to compression ratios of 8 through 40, which is the range likely to be used by most of the real-life applications permitting lossy compressi on. The proposed EC-PRVQ is found to give a good rate-distortion perfo rmance and clearly outperforms the state-of-the-art image compression algorithm developed by the Joint Photographic Experts Group (JPEG). Th e robustness of the EC-PRVQ is demonstrated by encoding several test i mages taken from outside the training data. The EC-PRVQ not only gives better performance than JPEG, at a manageable encoder complexity, but also retains the inherent simplicity of a VQ decoder.