Research Of Subjective/Objective Quality Assessment And Perceptual Optimized Coding For 3D Video

Currently, for normal home users, the experience of viewing 3D video online is not sat-isfactory, the image is blurry and the sense of 3D is poor. This is partly because of the huge amount of data of the 3D video. During online transmission, the quality of the video have to be reduced to save bandwidth. Since the sense of depth are greatly influenced by the quality of the 3D video, the realistic sense will be damaged due to image degradation. On the other hand, How to evaluate the quality of 3D video is still a problem. The distortion type of 3D video is different with that of the traditional 2D video, Directly applying existing video quality assessment method may not work, not to mention the user quality of experience of 3D video also contains more complex factors such as the sense of depth. This paper focuses on the sub-jective experiments methods and objective quality assessment algorithms for 3D video, and this paper also study how to integrate the objective quality assessment model into an existing 3D video coding framework. The depth map is now being widely used as 3D video auxiliary information. For depth map based 3D video system, the quality of the virtual viewpoint video will affect the user experience of quality of the whole 3D video system.The Chapter 2 mainly studied the subjective and objective video quality evaluation of the virtual synthesized videos which suffer from texture/depth compression distortion. The design principles of the subjective database is to ensure the quality of the testing synthesized videos will cover a wide range and have a distinction of quality between each testing videos. In order to achieve this goal, each texture/depth compression quantization parameter combina-tions are carefully carried out from a large number of selected candidates. The created virtual synthesized video quality database has been publicly released, and has been used by many well-known research institutions. In addition, the Chapter 2 also presents a full reference ob-jective quality assessment algorithm for virtual synthesized video. This algorithm focuses on the temporal flicker distortion induced by the depth map compression and rendering process itself. Experimental results show that the performance of the proposed algorithm is signif-icantly better than the existing objective video quality evaluation algorithm on the full data set, and the advantage respect to the other algorithms is more obvious on a subset which have obvious temporal flicker distortion.The Chapter 3 focuses on depth quality assessment of 3D video, and explore by sub-jective experiment on how image distortion affect the depth perception quality of 3D video, the experimental data set includes not only the symmetric distortion, but also the asymmetric distortion. The subjective experimental results prove that the loss of image details can affect depth perception. The scores of depth perception quality and picture quality obtained through subjective experiments have been publicly released. Meanwhile, the Chapter 3 also presents an objective depth perception quality evaluation algorithm for measuring depth quality degra-dation caused by image distortion. Experimental results show that the proposed algorithm can accurately predict the degradation of the depth perception quality.The Chapter 4 proposed a perceptual depth video coding algorithm. The proposed coding algorithm is based on a low complexity synthesized video quality assessment algorithm which can measure both the spatial distortion and the temporal flicker distortion. In this paper, the proposed low complexity algorithm is incorporated into rate distortion optimization process of depth video coding. Because the distortion metric has changed, the Lagrange multiplier of the rate-distortion optimization process was re-derived and re-adjusted. Experimental results show that, the proposed perceptual depth video coding algorithm can significantly reduce the temporal flicker distortion in virtual viewpoint video. The prediction accuracy of the proposed quality assessment algorithm is better than existing mainstream video/image quality evaluation algorithms. The quality assessment algorithm is simple and easy to be integrated into the existing 3D video encoder.