Research On Key Technology Of Video Coding Based On Human Visual System

In recent years, various novel video coding technologies have become more and more popular due to the requirements of multimedia services, such as scalable video coding (SVC) high efficiency video coding (HEVC), stereoscopic/multiview video coding (MVC) and distributed video coding (DVC). Since coding performance and robustness are also the most important issues in these coding technologies, a lot of researches have been attracted to study on them. However, most researches do not explore the visual perceptual properties. It is well known that HVS has different sensitivity of different video contents due to its underlying physiological and psychological mechanism. Thus, perceptual video coding is a new approach to improve the coding performance and robustness by exploiting the human visual system features. In this thesis, the perceptual stereoscopic/multiview video coding, perceptual distributed video coding and perceptual error-resilient video coding shemce are studied, respectively:Firstly, the thesis proposed a perceptual stereoscopic/multiview video coding based on adaptive residual preprocessing to achieve better coding performance. This scheme consists of two stages. To characterize the property of stereoscopic perception in auto-stereoscopic display, we utilize the best stereoscopic viewing position to establish a foveation weighting model, and present an image-domain stereoscopic JND model by integrating the foveation weighting model with the basic JND models to detect the stereoscopic perceptual redundancy. Then we propose a block-adaptive residue preprocessing method to reduce the unnecessary perceptual redundancy by minimizing the overall rate-distortion cost. Extensive experiment results demonstrate that the proposed scheme can efficiently reduce the unnecessary perceptual redundancy without visual quality degradation, especially for the high bitrates. In addition, the block-adaptive residue preprocessing method is also suitable for mono video coding. The simulation results confirm that our proposed method can achieve similar subjective quality with less bit-rate.Secondly, the thesis proposed a perceptual distributed video coding based on adaptive quantization to achieve better coding performance. The main contribution is the novel idea of introducing a perceptual distortion probability model to estimate the perceptual distortion of side information (SI) frame and the target perceptual distortion. We calculate the perceptual distortion probability for each DCT band of SI frame and conduct the target perceptual distortion probability by minimizing the RD cost. According to the above two perceptual distortion probabilities, adaptive quantization matrix can be determined for each frame. Extensive experiment results demonstrate that the proposed scheme can adaptively determine the quantization matrix to obtain similar visual quality but with less bit-rate by integrating the quality of SI frame, perceptual features and RD optimization.Thirdly, the thesis proposed a perceptual error-resilent video coding based on adaptive intra-update to improve the visual quality of compressed videos over packet-switched networks. Based on several observations of the content-dependency visual distortion propagation, we proposed two SSIM-based end-to-end distortion models to provide the estimation of the overall perceptual distortion. Then, a SSIM-based adaptive intra-update strategy is presented to maximize the visual quality of decoded videos for the given transmission conditions. Extensive experiment results demonstrate that the proposed scheme can achieve significant visual quality improvement for I1.264/AVC video coding over packet-switched networks by better preserving the structural information of the compressed videos.