Key Technology Research Of Three-Dimensional Display Based On Depth Cues

Since the movie "Avatar" was released in 2010,a 3D Upsurge has been set off among the whole world.However,the lack of 3D content has always been considered as one of the most serious problems in the 3D video display.In order to obtain 3D content,a number of methods have been proposed.In this paper,we achieve a kind of three-dimensional display which is based on a depth cue.In this paper,our target is to implement the three-dimensional display of 2D videos.We have deeply study the key technologies of the procedure,including the generation of depth map based on the depth cue,the generation of virtual viewpoints,the filling holes and the synthesis of virtual viewpoints to 3D.Firstly,we generated a depth map based on geometric cue and Grab cut matting.A depth map extraction method based on geometric information and Grab cut matting for the conversion of 2D to 3D is proposed.The process of geometric information estimation can be divided into two steps.First,Sobel operator and the probability Hough transform are adopted to detect the vanishing lines,vanishing point and level line.Second,the depth value of every point is assigned.Next,Grab cut algorithm is used to get the foreground from the original image.The foreground assigned by a same depth value is based on both the location of the foreground and the geometry-based depth map.Eventually,the geometry-based depth map and the foreground depth map are integrated into one by a depth fusion algorithm.Secondly,we generated virtual viewpoints using the technology of DIBR.In this paper,we use the current pixel value to add or to minus the times of its corresponding of depth value to generate the different virtual viewpoints.Then we must fill the holes in the virtual viewpoints.In this paper,we adopted LUO hole-filling algorithm based on Criminisi hole-filling algorithm to fill the holes.To improve efficiency,we apply one of push-down methods.That is to calculate which virtual viewpoint will be displayed in the current point,then just generate the virtual viewpoint and other virtual viewpoints will not be generated,improving the efficiency effectively.In this way,we achieved real-time three-dimensional display of the original 2D video.At last,as a meaningful application of three-dimensional display,super number virtual viewpoints naked eye display was referred to.We added technology of track based on the synthesis of virtual viewpoints.The entire implementation process can be divided into three steps:(1)Tracking the location of the viewer.(2)Calculating which virtual viewpoints will be displayed based on the location.(3)Synthesizing these virtual viewpoints.In this section,we achieved that a viewer can see different three-dimensional views in any two locations among the view range.