Intra mode selection method for depth maps of 3D video based on rendering distortion modeling

A technique to minimize distortions in synthesized virtual views, while encoding depth maps that are used in Depth Image Based Rendering (DIBR) applications is proposed. Depth maps are not viewed by end users, but are used for virtual view generation. Therefore, it is important to compress depth maps in a way that it minimizes distortion in views rendered with them. In doing so, it would be possible to generate high quality virtual views using compressed depth maps. Firstly, an error model to approximate rendering distortion caused by disparity changes is proposed. Thereafter, this error model is used at the encoding mode selection stage of coding depth maps. Experimental results illustrate an average bit rate saving of 19%-76%, compared with the mode selection method, which is based on minimizing pixel errors only of the depth map. Further, encoding depth maps with the proposed technique improves the overall visual quality of rendered views.