Related Techniques Optimization Of High Efficient Video Coding

3D video gives us a wonderful experience that we feel more real and involving compared with the traditional 2D video. With the gradual maturity of video display and imaging technique, 3D video coding becomes an attention topic to research institutions at home and abroad. Now the video format of the 3D video is texture plus depth, and the existing 3D video coding standard just transmits finite viewpoints with the texture plus depth. The video information of other viewpoints can be obtained by virtual viewpoints synthesis based on Depth Image Based Rendering(DIBR), which can save transmission bandwidth and memory space greatly. First, we give the optimization and improvement of the key techniques in view synthesis reference software(VSS), which is contained by the platform of 3D video coding standard(called RFD) proposed by audio video coding standard workgroup of China. Then, noise will be introduced in video information when coding, and the noise in depth map will lead to geometric distortion in synthetic viewpoints. So the quality of synthetic viewpoints’ image can be improved by depth filtering. According to the above content, related works are presented as follows:1. Optimization of hole filling in view synthesis: The hole filling method in VSS can not judge whether there is the background information around the hole accurately so that the hole will not be filled correctly. Using the foreground information to fill the hole will lead to quality reduction of synthetic viewpoints. In order to avoid the image distortion of synthetic viewpoints by filling the hole with the foreground information incorrectly, the background information that is discarded for disocclusion when 3D warping will be preserved in a mask, which is used to judge whether it exists the background information around the hole. If so, the existing background information will be used to fill the hole, if not, the preserved background information in the mask will be used.2. Improvement on edge detection and hole extension in view synthesis: In VSS, edge detection is used to mark the edge of the depth map with the width of two pixels. The pixels marked as the edge will not be warped because they are regarded as the inaccurate pixel points. So a lot of valid pixels in the edge are discarded, which causes the quality reduction of the image. The aim of hole extension is to eliminate the noise around the hole, but many valid pixels are cast off. To solve the problems above, single-point edge is proposed to mark the edge accurately and preserves more valid pixels in the edge. Simultaneously, the analysis and improvement are carried about the relation between the size of hole extension and performance.3. Depth spatial temporal view-reference filter: The reference pixel values and blocks can not be matched correctly in the depth spatial and temporal filter based on the relationship of 3D distance, texture and temporal correlation. So we propose a new method which contains block matching and view reference to filter the depth map. Block matching is more accurate than pixel vector matching that relies on the relationship of the pixel value of texture and depth of one pixel. In addition, different views have high relevance and it can offer the filtered pixels more precise reference pixels from the image of other views. The new method improves the quality of synthetic viewpoints.