The Key Techniques For Converting 2D To 3D Video Based On DIBR

In recent years, more and more three-dimensional video get into people's lives, bringing people the perception of three-dimensional immersive besides giving us the high-definition video and image. However, not only to capture the resources in traditional way is costly but also the capture equipment is complex. In order to transmit two ways of signals , network bandwidth must be very significant. The difference and the incompatible of display equipment of three-dimensional video and the current two-dimensional video . The technology of transferring two-dimensional video into three-dimensional video is based on image analysis, it means that we could get the three-dimensional video demonstration from the two-dimensional video directly. All of these hinder the development and popularization of three-dimensional video. And then the technology of Depth-Image Based Rendering(DIBR) is an effective method. More and more scholars also joined the study, proposing new questions and improving the method.This thesis first introduce the theory of three-dimensional vision , its principles and display device , then involve the format of two-dimensional video + depth image , in this way to obtain the three-dimensional video. In the article , we give a brief description of image based rendering , which have widely used in the multi-view video and stereoscopic video. Next, we described the method of Depth-Image Based Rendering in detail. Analyse the reason of the problems and solute them . We found that the disparity map which convert from the depth map have huge influence on the rendered image, especially about the hole ,so we propose the algorithm named Parallax-Map Base DIBR .In this way we could control the hole size before warping. After that we have study on the image inpainting , and use this method to deal with the hole problem. Of course we have done some improvement. Finally , we present our own framework of assessment ,the experimental results show that we have some improvements. The conclusions are made at last and some research topics along the lines of this thesis in the future and the possible hot problems are pointed out.