Human Visual System Based Rate Distortion Optimization Research

With the rapid development of computer and mobile communication, the using of multimedia technology, which core is digital image and digital video, is not limited in the broadcasting and television system. Multimedia technology is coming into everybody’s life.In the recent years, the people’s demand for video is increasing more and more. The high efficient video compression technology has become the basic core technology of home theater, digital broadcasting television, network television and so on. After one year’s study, JCT-VC released the new video coding standard--HEVC, which is constituted by the ITU-T Video Coding Experts Group (VCEG) and the ISO/IEC Moving Picture Experts Group (MPEG).According to the characteristics of HD/Ultra High Definition video, HEVC is added an amount of advanced coding technology, such as Coding Tree Units and Coding Tree Block structure, Advanced Motion Vector Prediction,4×4DST Transform, Motion Merge and so on. Under the premise of same subjective quality, HEVC saves34%-37%rate compared with H.264, which is easy to be applied in broadband network and mobile network.HEVC belongs to the hybrid coding scheme, which core theory is rate-distortion optimism. HEVC makes use of intra prediction to remove the spatial redundancy, inter prediction to remove the temporal redundancy, entropy coding to remove the statistical redundancy, transform coding to remove perceptual redundancy. To a certain extent, the Quantification technology removes a part of perceptual redundancy, but the study to perceptual redundancy is not perfect. Beacause the eyes is the final receptor of video and the human visual system has not the same ability to differentiate every region of the picture, it is essential to make use of human visual feature to modify the traditional distortion metric, which is to increase the compression efficiency.The proposed algorithm makes use of video saliency map and just noticeable distortion model to modify the distortion metric in order to make the subjective feeling and objective distortion metric equal in value. Finally, we adjust the Lagrange multiplier factor. The test results indicate that the proposed algorithm can adequately remove perceptual redundancy and improve compression efficiency in the condition that not reduce the subjective quality.