Perceptual Coding based on JND model for High Efficiency Video Coding

In this paper, we propose a video compression algorithm based on human visual perception for High Efficiency Video Coding (HEVC). The proposed algorithm is applied to the conventional lossy coding to reduce bitrate without any further subjective quality degradation. It improves coding performance by allowing objective quantization error which is not noticeable to human visual system (HVS). By using luminance just noticeable distortion (JND) model, the proposed algorithm finds out the maximum QP value based on JND threshold to reduce bitrate and not to degrade the subjective quality. Based on the JND threshold, the proposed algorithm quantize the transformed residual within a certain range of quantization parameter (QP) to minimize bitrate with retaining subjective quality in the HVS’ perspective. We found that the proposed algorithm shows maximum 18.37% and 8.27% in average bitrate reduction with similar subjective quality. Keywords-HEVC, perceptual coding, JND, Subjective quality