Efficient Depth-Based Coding

This chapter addresses predictive coding methods for depth maps that are required for virtual view synthesis. In multi-view immersive systems, virtual views play a very important role in the overall quality experienced by the users. Despite the fact that depth maps are not viewed by the users, their accuracy has a significant impact on the quality of the corresponding synthesized views. This is mostly due to the geometry information of the scene they represent, which enables reconstruction of any viewpoint lying between two camera views. The indirect impact of depth map quality on synthesized images and video suggests that the optimization of coding algorithms for view synthesis can improve the quality of experience of the user. In this context, the chapter discusses the compression of depth maps using standard coding techniques. A method for compressing the depth map with the use of the standard coding technique of advanced video coding is provided. A depth map quality metric is discussed as this is of paramount importance when allocating bitrates between the texture and the depth videos. Furthermore, experiments regarding compression efficiency and bitrate allocation strategy are described and their results presented.