The Research Of Efficient Depth Map Compression Focus On Quality Of View Synthesis In HEVC

With the development of3D display and interactive multimedia systems, new3Dvideo system is attracting significant interests. It could provide a more natural and livevisual experience in immersion and3D sense than the traditional2D display. However, new3D video brings multiply data and needs develop more efficient compression algorithm tosatisfy the new technology. As an important part of high efficiency video coding researchingin three-dimensional video, researching of high efficient depth map compression is studiedin this article. Analyzed the researching status and existing problems of the depth mapcompression, combine with the High Efficiency Video Coding standard, this paper has donesome research on efficient depth map compression focus on quality of view synthesis inHEVC. The major work of this paper is as follows:As the first step in this paper, research influence of distortion in different factor indepth map on distortion of virtual view rendered by depth map decode, and get threeconclusions that is the base of efficient depth map compression focus on quality of viewsynthesis. The conclusions are as follows: the bigger depth value produces the biggerdistortion in the same distortion intensity case; the samples distortion happened in textureregion have large influence on distortion of virtual view, but the samples distortionhappened in flat region have little; the appropriate smooth image can improve the quality ofvirtual view rendered by smoothed depth map.An algorithm of CU dec ision for depth map compression focus on quality of viewsynthesis is proposed based on the CU structure in HEVC. The algorithm consists of twosteps: firstly, partition the LCU region in depth with method statistics twice depth value.Secondly, make a CU decis ion based on divided LCU region under the guidance of thetheory that influence of different distortion in depth map on distortion of virtual view.Experimental results show a significant improvement of the initial input depth maps andcoding efficiency, as well as a good view synthesis quality.Considering depth map’s feature and intra prediction process in HEVC, a fast intracompression method focus on quality of view synthesis in HEVC is proposed. Proposedmethod is implemented with three steps in consideration of balance between compressionaccuracy and computation complexity. Firstly, utilizing depth map’s feature partition theLCU region simply. Secondly, make a CU and PB decision under the guidance of the theorythat influence of different distortion in depth map on distortion of virtual view. Finally,make a intra mode previous selection based on the divided PB region. Experimental resultsdemonstrate that the proposed algorithm can reduce the computational complexity morethan35%for every view, without decreasing on the quality of view synthesis.