Font Size: a A A

Research On Hole Filling In Depth-Image-Based Rendering

Posted on:2016-08-30Degree:MasterType:Thesis
Country:ChinaCandidate:Y DingFull Text:PDF
GTID:2308330479990251Subject:Information and Communication Engineering
Abstract/Summary:PDF Full Text Request
With the advance of displaying technology, the demand for visual enjoyment grows rapidly. The appearence of 3D movie provides people with the experience that they were at the scene, in the meanwhile, with the development of 3D industry, the big data caused by two views in 3DTV brings a chanllenge to the transmission system. Moreover,with the emerge of MTV and FTV, it becomes impossible to transmit all the viewpoint scenes. Transmitting one or two viewpoint and rendering the others at the receiver is a feasible solut ion to the problem. Depth Image Based Rendering is one of the most popular technique used in rendering new viewpoint scenses.This thesis makes a study on on the properties of DIBR, especially on the handling of holes caused by disocclus ion. To recover the texture in the hole area and raise the quality of rendered image, some research is done on the filling of holes in virtual viewpoint image.The work of this thesis is listed below: Firstly, this thesis inroduced the principles of DIBR technique and analysed the basic process of DIBR from the scope of C omputer Graphics. By introducing the the model of pinhole camera and the parameters of the camera, the 3D warping equation is deduced. Then the simplified DIBR is introduced and the problems in the process of DIBR is analysed. The problems such as overlapping, ghost contour and disocclusion holes. The settlement of overlapping and ghost contour is also given.Secondly, an algrithom called C riminisi is applied in the hole filling of virtual viewpoint image. A solution to the handling of hole filling in video is proposed. To speed up the algritho m, methods of background information abstraction and updating, small holes prefilling, subsampling and minimizing the searching area are used. Given the specificity of virtual viewpoint image, depth information is introduced in the process of both priority estimation and the best patch search. K-means classification is used when filling the depth map to avoid blob.Thirdly, permeability filter is used in the filling of holes caused by disocclusion. To avoid the stripe and the luminance discrepancy, histogram specification and median filter is used.Finally, to test the quality of the obtained 3D video, a 3D player in android system is designed based on HTC Evo 3D smart mobile phone.
Keywords/Search Tags:DIBR, hole filling, image inpainting, permeability filter, 3D player
PDF Full Text Request
Related items