21-3 適応予測による自然画像の可逆符号化についての検討

抄録

JPEG Spatial mode is known as reversible coding scheme for natural images. This scheme is based on Differential Pulse Code Modulation(DPCM), however that is not so effective because of fixed prediction. In this paper, we propose a reversible coding method for natural images. Fixed or adaptive prediction is used according to states of absolute difference value among neighboring pixels in our method. The prediction equation which gives the minimum entropy of prediction errors is used in fixed prediction mode, where as the prediction equation which gives the minimum power of prediction errors at the neighboring pixels is used in the adaptive prediction mode. Simulation results for the images of 256 levels show that the proposed method can reduce the entropy by an average of 5.73 percents compared with the JPEG Spatial mode.