A Layered Method For Virtual View Synthesis

With the fast development of radio and television related technology, the traditional two-dimensional display technology is no longer able to meet the demand of people for stereoscopic sensation and immersion. True 3D volumetric display technology enables the viewer to get rid of the shackles of external devices and regional limitation. This technology can let peo-ple produce immersive feel, and gives viewers an unprecedented immersive experience. True three-dimensional display technology is the ultimate goal of the three-dimensional display, is considered the most viable and will eventually become a common platform for the next genera-tion of display technology,and contains a huge industrial development opportunities, it has been widespread concern.True 3D volumetric display technology is the key to draw a virtual viewpoint, and in the current virtual view rendering techniques exist in the occlusions, resamplings, holes, artifacts, and efficiency is not high. In order to solve these problems, a layered method for virtual view synthesis is presented and implemented, which is based on the conventional virtual view image based rendering technology. The pre-processing of depth map is firstly used to solve the problem of inaccurate depth map, then non-boundary area, foreground boundary area and background boundary area are extracted using layered method, next 3D image warping and blending method are used to process this layers separately, finally in order to improve the image quality of virtual view image, the method of inverse warping is used to fill cracks, the filling method base on the depth information is used to fill holes, and virtual view image generated finally. In order to improve virtual view image rendering speed, we use the GPU to accelerate some of these modules process.The test data set of Breakdancers provided by the interactive Visual Media Group of Mi-crosoft Research is used for simulation experiments. And the experimental results show that the method mentioned in the paper can effectively solve the problem of occlusions, resamplings, holes and artifacts. This method have greatly improved in terms of speed and can be used to create the virtual view image in three-dimensional television.