Improving HEVC Coding Perceptual Quality Using Saliency Guided CTU Layer QP Adjustment

Recently, improving the perceptual quality of video coding under bit rate constraint has attracted researchers' attention in video coding field. In this paper, we propose to improve the coding perceptual quality for the latest High Efficiency Video Coding (HEVC) standard using saliency guided coding tree unit (CTU) level quantization parameter (QP) adjustment scheme. Our work mainly includes two steps: first, a CTU layer saliency detection scheme with low complexity is proposed to measure the visual significance for each CTU in a frame. Then, a CTU layer QP adjustment scheme based on relative saliency is proposed to increase the bit rate resource for more salient regions and decrease the bit rate cost for less salient regions. Therefore, the proposed QP adjustment scheme obtains better bit rate allocation among CTUs and generates better coding perceptual quality. Experimental results show that the proposed method achieves better perceptual quality and rate distortion performance for HEVC video coding.