Research Of3D Video Coding Based On Hevc

The improvement of3D video technology such as3D movies,3D broadcast channels and Auto-stereoscope displays has led to a growing interest in3D video research. The Motion Picture Experts Group (MPEG) has been conducting its international standardization activities of FTV (free-viewpoint TV) since2001, which is a framework that allows viewing of a3D world by freely changing the viewpoint. Multi-view Video Coding (MVC) is the first phase of FTV, and3DV is the second phase of FTV. The Joint Collaborative Team on3D Video Coding Extension Development (JCT-3V) from ITU-T Study Group16(VCEG) and ISO/IEC JTC1/SC29/WG11(MPEG) has been created to develop3D video coding technology, which has reviewed the work previously done on3DV coding of MPEG and has established the test model of the AVC3D extension framework (3D-AVC), Multi-view HEVC extension (MV-HEVC) and HEVC3D extension framework (3D-HEVC)In3D-HEVC framework,3D video is represented by the Multi-view Video plus Depth (MVD) format. At the encoder, inter-component dependencies between texture and depth are exploited and joint coding of texture and depth data is introduced. In the independent view, also called the base view, an unmodified HEVC codec is used. And for the coding of dependent views, additional tools have been integrated into the HEVC codec. The work of this paper based on3D-HEVC.Firstly, the base framework and key techniques of3-HEVC are discussed and three difficult research point of3D video coding are included:texture video coding, depth map coding and depth estimation.Second, fast texture image mode decision algorithm based on depth map on the3D-HEVC framework is introduced. This algorithm exploits the relationship of depth map and texture image in3D video, dividing the texture image foreground scene, main scene and background scene based on the depth value. And the relationship of depth value and the motion vector are introduced to speed up the mode decision between inter-view and inter-frame.Thirdly, the depth map coding of3D-HEVC is introduced. Based on the feature of large-scale smooth region and sharp edge in depth map, four new predict modes are introduced on the inner-frame prediction of depth map. And based on the feature that the depth map is for rending the virtual view, the new mode decision function is introduced, called view synthesis optimization for depth data.The experiment section of this paper conduct on the HEVC based3D video coding,3D-HEVC. First, coding efficiency is under comparison between3D-HEVC and simulcast. Secondly, depth map based fast mode decision algorithm for texture image on3D-HEVC is verified which speed up29.3%compared to original software on texture coding. A bitstream analysis software of H.265/3D-HEVC has been developed under Linux with the help of QT and HM.