Research On High Efficiency Intra Prediction For Texture And Depth Maps Coding

With the rapid development of information technologies, the requirements of information with various contents, for example, text, digital audio, image and video information, grow explosively. Among all these information, digital video could give vivid and delightful perception to users, and the percentage of digital video information in all these information is up to70%, therefore, digital video technologies has been concerned a lot by the industry.Digital videos could be divided into two dimensional (2D) and three dimensional (3D) videos.2D videos consist of a series of temporal successive pictures, and3D videos consist of a series of2D videos in different view directions. Besides,3D videos contain depth information of all pixels in the series of2D videos. In order to compatible with the existing video coding standards,3D videos were mainly represented by multi-view video plus depth maps technology.The biggest challenge of video technologies is to transmit the tremendous volume of data via the available networks, thus, the video compressed technology is critical. H.264/AVC can give higher coding efficiency than any other video coding standards because it provides many complex modes. The research of project on the new standardization for high efficiency video coding (HEVC) is under thorough discussion, which was focused on improving the performance of H.264/AVC and3D video coding.Intra prediction was the core module of H.264/AVC with high computation complexity, because it contains many complex modes, so the research of improving the performance of intra prediction mode is critical to H.264/AVC. The main contributions of this thesis are high efficiency intra frame coding. For texture video, the Grey system theory is employed for H.264/AVC intra prediction coding, for3D video coding, a planar model based intra prediction method for depth map is proposed. The principal contributions of this thesis are concerned with the following aspects: (1) A Grey Model Application Method for H.264/AVC Intra Prediction CodingGrey system theory is employed to solve problems with less data, little sample and insufficient experience, it has been successfully applied to medicine, industry technology and so on. In this paper, a Grey Model (GM) based intra prediction method for H.264/AVC is proposed. For a certain block, when all the predicted values (obtained from the existing intra prediction methods) are similar, they are sent to a Grey system so as to obtain the final prediction results. Accordingly, the rate distortion decision procedure could be saved; meanwhile the mode information could also be saved. Since the encoder and decoder can determine whether different prediction results are similar or not, no additional flag information should be included into the bit stream. Experimental results demonstrate that by integrating the GM based prediction method into the H.264/AVC Joint Model version15.1(JM15.1), an average2.994%bit rate and39.28%coding time can be saved while maintaining the same quality of reconstructed videos.(2) A Planar Model Based Intra Prediction Method for Depth Map CodingA planar model based intra prediction method for depth map is proposed based on the analyses of the nature of depth maps. The neighbouring pixels as well as their corresponding pixel coordinates of current block are utilized to compute model parameters. Then, the predicted pixels of the current block are calculated based on the proposed model. Finally, a rate distortion criterion is used to choose the optimal prediction method between the proposed one and those of H.264. Experimental results demonstrate that by implementing the proposed method into H.264/multi-view video coding (MVC) software platform Joint Model of Multi-view Video Coding version8.3(JMVC8.3), the maximum and average bit rate savings of depth maps can be achieved as high as-13.14%and-6.35%, respectively, while maintaining the same quality of synthesized virtual views.Experimental results show that the two proposed methods can obtain satisfactory performance in intra prediction for2D videos and3D videos respectively. They both reduced the computational complexity and improved the perfonnance of H.264/AVC.