Research In2D To3D Moving Images Conversion Technology And Its Multi-Coreparallel

In recent years,3D movies are increasingly popular, However, due to the3D video s highercost and longer production cycle, there are not enough3D video sources for people to watch. Nowthere are many traditional2D videoes are used widely, so to resolve the question of the lack of3Dvedio content, many people begin to study the2D to3D video conversion technology.In this paper, firstly, we introduce the theoretical and konwledge of the three-dimensionalvedio. And then a depth map extraction method based on motion and texture characteristics isproposed, for the characteristics of most iamge sequences exist the relative movement of theforeground and the relatively stationary of the background. Separating the foreground andbackground, and extracting the depth information respectively, then combining them effectively.Because this method combines the different depth cues, the depth information is more closer to thereal scene.Secondly, we introduce the techniques of Stereoscopic image rendering, and research thetechnology of the2D to3D conversion technology, propose an method which combined themathematical morphology and the linear interpolation to fill the holes. The results of theexperiments show that can get the image sequence of high quality by using the proposed method.Finally, because the block-matching algorithm has the characteristics of good parallelism, theGPU-based parallel global block-matching algorithm is proposed, which can be more faster thanthe treditional CPU implementation.