The Research Of Multi-view Video Coding And Viewpoint Rendering

In recent years, digital multi-media have tremendous development，On the one hand, the technology becomes more and more mature in the fields of computer vision and computer graphics，On the other hand, computing devices and image processing technology were improved,so that more media type were developed,applied,spread.In consumption area,3D video has become a bright spot in digital multi-media area.Compared with single video,multi-view video contains large volume of information which increases with increasing amount of viewpoint.The large volume of information bring major hardships for storage and transmission.So that Multiview Video Coding(MVC) was developed to minimize the data size,and became a key technology of3D video.Since the multi-view videos were taken by shoting the same scene from different angles,there is correlation between viewpoints.MVC reduce the redundant data between viewpoints,using the correlation and combining the traditional video compression technology,in order to compressed video data.Besides compressing video data.viewpoint synthesis technology is also a key technology of MVC. Viewpoint synthesis technology is based on the correlation between viewpoints.using this technology we can synthesize any viewpoint video.According the placing of cameras and the various parameters of camera,viewpoint synthesis technology can process the image and disparity information to synthesize virtual image.In many application,multi-view video is represented by video plus depth.The depth value describing the depth of objects in the scene.This structure has advantages.At the decoded side.based on video plus depth,using3D warping,we can synthesize virtual view.In this paper, we analyzed the geometric theory of3D mapping in DIBR technology and the key problems such as occlusion, the pixel overlap, hole and so on. On the basis of the existing methods, proposes a combination of depth map pretreatment and interpolation to fill the empty method to synthesis image to fill empty. Experiments show that, this method can effectively remove the empty graph of synthesis of, and achieved good visual effect.