| Depth video plays an important role in 3D video systems based on multiview plus depth(MVD) format. With depth video coded in addition to the texture video, 3D video system can archive arbitrary viewpoint by using depth image based rendering(DIBR) which provides 3D perception efficiently and has low processing cost. During the video coding process, coding artifacts in depth video could cause degradation of virtual view. A new depth coding method should be proposed depend on its own characteristics. In addition, as depth video and corresponding texture video are synchronously acquired from the same scene in MVD format, they are highly interrelated. Thus, depth video can be compressed using the coded texture information.In this paper, a region based trilateral filter for depth video coding is proposed to improve the coding quality. First, depth map is split into multiple objects with the label of segmentation. Then, in order to remove the artifacts from different objects and preserve the sharp edge, each pixel is filtered by using neighboring pixels belong to the same object. The proposed method has been evaluated with several 3D video sequences in HEVC platform. Experimental results demonstrate that the video quality can achieve a considerable gain under the same bitrate.With respect to the interrelation between the texture video and the corresponding depth video, a texture assisted disparity vector prediction method for multiview depth video coding is proposed. First, the co-located texture block for current depth block is located and the disparity vector of the block is derived as a candidate of current depth block. Then, another disparity vector candidate is obtained with depth information. These two disparity vectors are compared by calculating down-sampling mean square deviation, and a more accurate one is selected as the disparity prediction of current depth block. Experimental results show that the proposed method can achieve higher performance compared with HTM encoder. |
PDF Full Text Request