Research On Binocular Images Based Methords Of Image Convertion From2D To3D

In recent years,3D storm have been sweeping the world. The number of peoplewho are keen on watching3D videos are becoming larger and larger, in themeanwhile,3D TVsets are also coming into many ordinary families. With thedevelopment of3D industry and the expanding of market share, the scarcity of3Dvideo resources become an obstacle to3D technology. Many researchers arethrowing there eyesights on the field of3D at home and abroad.Thinking about the generation of3D video, it belongs to the area of imageprocessing. The key point of the tramsform from2D contents to3D ones is to obtainthe depth map. After a comprehensive analysis of traditional methods, the researchis based on binocular images. In this paper, two improved disparity map acquisitionmethods are shown, and then accomplish the composite of3D images.Firstly, the thoughts of image segmentation and belief propagation algorithmare combined to achieve the accurate disparity map. The results show that,compared with pixel based belief propagation algorithm, it reduces the timeconsumption and is more efficiency. At the same time, it enhances the robustness ofthe algorithm, meanwhile, the sensitivity to the distort point is significantly lower.For the most important, the more accurate disparity maps are generated.Secondly, the model of optical flow field is introduced into the research toobtain the depth map from binocular images. For finding the differences of opticalflow field in the binocular images, the methods for motion analysis is referenced to.In order to obtain more accurate depth maps, the concept of belief factors isintroduced to the process of optimization. The results show that the parallaxinformation obtained by optical flow method has been optimized, edges betweendifferent regions are more clear. It also makes it smoother in gradation regions. In aword, the depth information of the object from binocular images is more accurate.Finally, in order to do further tests, a convertion from2D image to3D image isnecessary. And then, the subjective evaluation by human beings to judge theaccuracy of the depth map generated can be carried out. The technology calledDIBR based on human visual characteristics is used to synthesized the image fromleft view and right view. So that,3D image can be synthesized. The outcomes showthat the3D achievements have been improved in varying degrees.