Signal Processing: Image Communication

The multiview video exploits both texture and depth video information from various angles to create a 3D video [1]-[3] and free viewpoint video (FVV) [4] which are gradually becoming more popular for their advanced visual experience with depth perception [5]-[7]. Unlike texture, depth video is determined by a gray scale map indicating distance between camera and 3D points in a scene [8]. If a comparison is drawn between depth and texture image coding, the former one incurs with extensive burden in terms of detecting and encoding its complex edges [9]. Standing on some texture-depth motion correlations, a number of methods in literature encode both texture and depth videos jointly by using texture motion information for the corresponding depth coding. The intention is to reduce the coding complexity by completely/partially avoiding the costly motion estimation process. However, those approaches suffer from the following inescapable limitation: the texture similarity metric is not always equivalent to the corresponding depth similarity metric especially at edge levels. Therefore, those coding techniques could not explicitly detect and encode the complex edge motions of depth objects and reach the similar or improved rate distortion (RD) performance of the HEVC reference test model (HM) [10].