A Perceptual Rate-Distortion Based Quantization Level Adjustment Algorithm

It is desired to obtain the best possible image quality subject a predefined bit rate in video coding. Aiming on improving quantization performance for signals in non-flat region, the fast rate-distortion based quantization level adjustment algorithm (FRDQLA) was proposed in our former work. It can improve the objective coding performance efficiently. In this paper, a perceptual FRDQLA algorithm is proposed. By introducing the HVS model, the perceptual weighted rate-distortion cost is calculated. The signal is quantized, coupled with an advanced vision model based perceptual distortion measure for rate-distortion optimization to produce coded images with improved visual quality. The proposed algorithm has no overhead and is completely compatible with the standard decoder. Experimental results show that the proposed algorithm can efficiently improve the subjective viewing quality and reduce the bit rate 11%.