A prequantizer with the human visual effect for the DPCM

Abstract

Abstract In this paper, a new differential pulse code modulation (DPCM) structure is proposed to improve the conventional DPCM. In the conventional DPCM, the design of the prediction error quantizer is mainly a nonuniform optimum quantizer. This kind of the optimum quantizer must require some statistical data in advance such that the complexity of the entire system increases. In this paper, the proposed DPCM uses a prequantizer to prequantize image data instead of an optimum quantizer to quantize the prediction error. The prequantizer only requires the mean of the image as a parameter to adapt the quantizer. In addition, the quantizer also matches the human visual effect, the Weber's Law. The prequantizer increases the correlation between pixels largely such that the predictor can predict the next pixel more accurately. It reduces the entropy and bit-rate of the prediction error largely. The simulation indicates the bit-rate with the proposed method is 1 bit per pixel less than that with the conventional DPCM under the same image quality. The proposed method also compares with the block truncation coding (BTC). Under the same bit-rate, the image quality with the proposed method is 10 dB better than that with the BTC method. In addition, the computation complexity of the proposed method is not greater than that of the conventional DPCM or BTC.