An adaptive image coding algorithm based on noncausal LPC and vector quantization

An experimental study is reported of a new adaptive monochrome image coding algorithm based on a noncausal linear predictive model and vector quantization. The input image is partitioned into blocks in this algorithm, and a scalar quantizer is designed for the sample mean. Adaptation is achieved in two stages. First, 2-D noncausal linear-predictive-coding (LPC) analysis is performed on the image blocks and the resulting linear predictors are vector quantized. Then the residual signals are coded with an adaptive vector quantizer which uses the spectrum and residual energy information from the LPC codes to control bit allocation. Experimental results show that good coding quality can be achieved at low bit rates