Rate distortion performance of VQ and PVQ compression algorithms

Summary form only given. This paper investigates the performance for memoryless sources and sources with memory by using vector quantization to encode and decode the source data. By modeling an image to be a Markov source, the authors suggest a lower bound estimate of the rate-distortion function for the image with memory which can be used to evaluate the performance of VQ (Vector Quantization) and predictive VQ. For the latter, the residual image, which is the difference between the original and the predictive image, is used to generate codebooks. In the encoder, three previous encoded pixels are used to predict the current pixels. The errors between a codevector and the corresponding predictive vector are compared and the minimum criterion is used to select the codevector