Research On View Rendering And Display For Natural Three-dimensional Television

The auto-stereoscopic display presents to each observer stereoscopic pairs of images and provides a strong binocular disparity and 3D illusion, making observers feel right on the scene and is sure to be the next direction of display technology. The natural 3DTV system dedicates to capture of natural scenes, backwards compatibility and providing depth sense without burden. Research on natural 3DTV is of far-reaching significance.Compared to other 3D expression formats, "2D video plus depth" has the advantages of high compression rate,backward compatibility and terminal flexibility, which makes it become the research direction and hotspot. This paper makes research on the field of rendering method, disocclusion restoration and multi-view auto-stereoscopic display technology based on "2D video plus depth" format. The achievements are as follows:(1) A new virtual view rendering method aimed at flat-panel auto-stereoscopic displays is proposed that shifts each pixel in reference image horizontally based on depth image. The method pre-processes the depth image by expanding objects boundaries so as to avoid foreground being split due to the inaccuracy of depth image. All points in reference image are shifted horizontally to the desired virtual position based on associated disparity, which is determined by depth image and the limit of screen parallax. Compared with traditional DIBR method, it is widely used since camera calibration parameters are not necessary. In addition, this method can control the depth sense according to individual preferences, giving each audience comfortable and stunning 3D illusion.(2) The paper proposes a depth-aided time-spatial consistency inpainting algorithm for disocclusion restoration after image warping. The method first excavates information from image sequence on the direction of time to repair part of holes, then repairs the remains with improved depth-aided inpainting algorithm. In this algorithm, the restoration order is based on the block priority that prefers to the block close to the background. When choosing the most matching block, this algorithm selects the background texture as a priority. Compared with other methods, proposed method has the advantage of full use of the information on the spatial and temporal direction. It has the ability to provide virtual view image with imperceptible signs of repair.(3) A view-weighted composition method for the most popular lenticular-based auto-stereoscopic display is presented. The method is made up of three parts, i.e. RGB component mapping to determine the view number of a given point in the plane of the LCD, sampling for each view through binocular interpolation, and RGB component arrangement using view weighting. The experiment results are quite satisfactory in both synthesizing efficiency and viewing effect.