"Advanced Keying, Compression and Watermarking of 3D Video Signals"

点击翻译按钮获取中文摘要

3D TV programs are an irreversible trend for future TV broadcasting. Research on 3D video processing has become an extremely hot topic in recent years. We, as researchers in the field of video processing, must get ready and prepared for the immediate future of video processing. We should research and develop advanced algorithms to produce, process, store, and transmit 3D TV contents. In this project proposal, we will focus our research on keying, compression and watermarking of 3D video signals. The research results can be extended to multiview 3D video processing. Chroma keying is a technique for compositing two images or frames together in which background colors from one image is removed, revealing another image behind it. To the best of our knowledge, there is no keying algorithm that can process semi-transparency sufficiently well. Applying the technique to video signal is even harder since extracted mattes on sequential frames must be both spatially and temporally coherent. We will research and develop advanced real-time 2D/3D video keying algorithms by using depth information in addition to chroma, spatial and temporal information. Depth map generation is indispensable for 2D to 3D video conversion and 3D video representation using depth maps. We will research and develop advanced algorithms to generate depth map for 2D/3D video signals, and SSIM based quality evaluation algorithms for keying and depth map. We will use the new quality metric to guide the generation process of keying and depth map. The 3D video representation using depth maps (called 2D plus depth) will become more efficient, which represents the stereoscopic information using the monocular image and its associated depth information. We will research and develop advanced depth map compression algorithms for 3D video compression. The proposed research on 3D video keying, compression and watermarking will definitely benefit the video industry of Canada.