Shape-Based Depth Map Coding

Three-dimensional television (3DTV) system based on depth-image-based rendering (DIBR) has won popular recognition of researchers. Since the depth information is closely related with the rendering of viewed pixels, the quality of coding has a direct impact on the later effect of the three-dimensional(3D) video, of which the existence of cloud noise and the self-adaption adjustment of depth have become the major problem of the current 3DTV system. By deeply understanding the characteristics of depth information´s object commonality and effective coding of depth information according to multi-objects, this paper proposes a new three-dimensional data representation ldquotwo-dimensional (2D) texture + depth map + shape masksrdquo. The proposed algorithm establishes its model according to the similarity of different objects´ depth information. This algorithm can not only effectively erase the cloud noise caused by the latter synthesizing of the uncertain boundaries, but also conduct the self-adaption adjustment of the depth to the practical needs. It exhibits great advantage in performance compared with such coding depth maps as MPEG-4.