Research On Depth Maps Creation For3DTV

With the development of television technology, the traditional2D television can not satisfy people’s demand for an immiscible experience watching TV. So Three-Dimensional Television (3DTV) becomes a hot research topic in resent years.3DTV is considered to be the next generation after high definition television (HDTV) in the television history. Therefore, it has great theoretical and practical significance to research the technologies of3DTV.The paper discusses several classical Three-Dimensional Television systems, as well as the research status of three-dimensional display, coding techniques. The key problem of three-dimensional television system is depth map creation, so the paper research on depth map creation algorithms. Depth map creation algorithms can be divided into two methods, relying on the depth camera and relying on the stereo matching algorithm. Due to the higher cost of the depth camera and its limitations, this paper focuses on the research of stereo matching algorithm.Stereo matching is the core component of computer vision, which is widely used in the practice of3DTV system, automotive driver assistance, and computer vision.3DTV system requires accurate depth map, and if the television program need to be broadcast live, real-time nature of the algorithm is also very important. So this paper do a lot of research on the stereo matching algorithm, and presents a fast and accurate stereo matching algorithm, which is based on image segmentation. This algorithm is based on the framework of the Tao algorithm and the whole algorithm includes several steps, such as image segmentation, initial disparity acquirement, initial plane fitting, region and plane refinement, and plane labeling. In the image segmentation step, we use the Mean Shift segmentation algorithm which can get accurate and fast segmentation; In the initial disparity acquisition steps, the cross-aggregation algorithm is used, which is an accurate and rapid local stereo matching algorithm; The region and plane refinement makes our algorithm more robust; In the plane labeling step, in order to further improve the real-time nature a new evaluation function is used which is different from the general Tao framework function. Experimental results show that this is an accurate and fast stereo matching method.