APPLICATION OF ITERATIVE ALGORITHMS TO ADAPTIVE PREDICTIVE CODING

Application of various iterative algorithms to adaptive predictive coding (APC) of image data is described. The variable steepsize steepest descent (VSSD) algorithm, obtained by coupling the adaptive predictor to the adaptive quantizer (Mark, 1976), represents the best candidate in terms of system performance and system complexity. The Kalman algorithm exhibits fast convergence and good distortion measure, but is too complex for implementation. For grey level image coding, an adaptive Laplacian quantizer in conjunction with adaptive predictive coding exhibits superior performance to an adaptive Gaussian quantizer. APC coding of image data offers at least a 7dB improvement in SNR over optimum PCM at any quantization level.