Depth boundary reconstruction based on similarity of adjacent pixels for efficient 3-D video coding

This paper proposes an efficient technique to improve 3-D video coding performance by taking into account coding artifacts and spatial resolution. Although a conventional video coding standard is used to compress the 3-D video such as multi-view plus depth data, it often introduces serious coding artifacts around object boundaries where signal abruptly changes in color and depth images. In this paper, to suppress the coding artifacts around object boundaries in compressed depth map, a depth boundary reconstruction method is proposed by using the similarity between a target pixel and its adjacent pixels. Since the proposed method provides the improved quality of synthesized view, it is used as in-loop for improving 3-D video coding performance. In addition, the depth down/up-sampling coding approaches are used to further reduce the bit-rates. To verify the excellence of the proposed method, it was applied to compress the multi-view plus depth data which is well-known as an efficient data format for representing the 3-D video. Experimental results show that the proposed method is helpful for improving the 3-D video coding performance and obtaining the quality of synthesized view in terms of both objective and subjective measures.