DPCM Picture Coding with Adaptive Prediction

Adaptive prediction is a method of improving the prediction in differential pulse-code modulation (DPCM) systems. The information on contour directions derived from neighboring picture elements is used to select a suitable prediction value for the actual sample. An effective way for determining the contour direction, taking into account signal-source noise, quantization distortions and hardware complexity, is given. Adaptive prediction can be favorably combined with a switched quantizer. Simulations both with critical test signals and broadband TV pictures have shown that adaptive prediction can substantially improve the quality of pictures with high-definition contours. The sensitivity to signal-source noise and to transmission errors is small and comparable to fixed prediction DPCM systems.