| Recently, with the showing of many excellent3D movies, three-dimensional television (3DTV) has drawn people’s attention. TV technology has experienced black-and-white television, color television and high-definition digital television, and now it has entered a new era of3DTV.3DTV can display the scene more realistic and can give the audience immersive felling.3DTV technology can be used in entertainment, advertising, arts and other areas. Now many Research institutions all over the world are studying on3DTV technology.This paper introduces the development and current situation of3DTV, and discusses the important part of3DTV system, including program source production, coding technology and three-dimensional display. Because the3DTV system needs to transport multi-points views, we usually use the coding mode of "video information+depth information" to reduce the bandwidth and use DIBR technology to draw the views in the receiving terminal. The extraction of depth map is very important in3DTV system. We can get depth map through the depth camera or stereo matching algorithm. The depth camera is very expensive and has a low accuracy. So we use the stereo matching algorithm to get the depth map.The stereo matching algorithm can get disparity map, which can be transformed to depth map within known camera parameters. According to generated depth maps, the stereo matching can divided into sparse depth map generation algorithm and dense depth map generation algorithm. The sparse depth map generation algorithm includes Feature-based stereo matching algorithm. The dense depth map generation algorithm includes area-based stereo matching algorithm and global-based stereo matching algorithm. Because in the receiving terminal of3DTV system, the DIBR technology need dense depth map to draw the views, so we can only use dense depth map generation algorithm, including area-based stereo matching algorithm and global-based stereo matching algorithm.3DTV system requires high accuracy and also need real-time display. This paper proposed a segment based stereo matching algorithm which has good performance in accuracy and speed. In image segmentation step, we use Mean Shift algorithm to has a good performance in segmentation. Then we use area-based stereo matching algorithm to get the initial disparity. In order to remove the unbelievable points, we do a cross checking process. Then we allocate the disparity to every image segment. Experimental results show that this method has a good performance in accuracy and speed. |
