研究期間：10208~10307;When the video in the cloud provides the clients 3D video (3DV) services such as free viewpoint videos, the huge amount of 3DV data can be much reduced by data compression. Accordingly, the cost of service provider can be much decreased while the clients are provided with high-quality free viewpoint videos. On the other hand, to provide efficient compression of 3DV and high-quality free viewpoint video, the international standard organization, MPEG, initiates the final call for proposals in March 2011, where the multi-view video plus depth (MVD) is taken as one of the upcoming 3D video formats since it enables a receiver to generate virtual views. In addition to view synthesis, the technology for depth estimation also has a significant impact on the quality of reconstructed views. Motivated by this, the main goal of this project is stated as follows. First of all, this project will enhance accuracy and reliability of occlusion map from 3D videos. This will improve the quality of depth videos and assist the high efficiency 3D video coding for view synthesis. Next, we will focus on speed up the proposed occlusion detection, disparity estimation, and high efficiency 3D video coding. With this, the visual quality of the synthesized virtual view can be improved with significant decrease of bitrate. With these, the high-quality synthesized views will provide viewers good entertaining experiences.
