Key Technology Research Of Stereo Video System Based On Depthmap

Stereo video system provides us a real sense of distance and makes us feel that we can touch the object in the screen. However, the expensive equipment, the complicated manufacturing process and the long period it takes to obtain a video obstacle the volume production of stereo videos. Stereo video system turns2d videos into3d ones using technology of computer vision, image processing and stereo technique. Such systems satisfy people’s need in3d movies and arouse more and more attentionThe paper designs a system with a fixed video format,2d plus depth. The paper does research in the relevant algorithms such as, stereo matching and depth based image rendering (DIBR). The paper also gives a new DIBR algorithms based on foreground object protecting.The paper realizes two different kinds of stereo matching algorithms. One is Mini-Census algorithm based on local region matching. The other one is Graph Cuts algorithm based on whole region matching. The GC algorithm’s result shows better quality after our experiment but it will not be used in our stereo system because of large computational complexity.The paper does research in DIBR algorithms, and also realizes and modifies a DIBR algorithm based on foreground object protecting. Before the depth map smooth process, the proposed method performs a depth map registration in order to reconcile the depth map with the color image. Then the algorithm uses a piecewise filter to do smoothing only in the background region of depth map. The proposed method visually protects the foreground objects in generated right view images. Moreover, it has a better PSNR performance and relatively smaller hole size compared to previous works. Better stereoscopic quality and real-time DIBR could be achieved by using this method.After the research of key algorithms, the paper builds a stereo video system using all of those technologies. The system confirms the practicality of those algorithms.