Research On Grab Cut Matting Algoirthm In The2D-3D Conversion

With the3D version of the film "Avatar" showed，People’s interest in theThree-Dimensional Display System has been increasing phenomenally. For lack of3D material, to Dig out the objects that people interested from the2D resources bymeans of Grab cut which as a2D to3D conversion technology in3D source costrelatively low and the production is also more flexible.Grab cut is a relatively mature of interactive color Image Segmentationalgorithm which has developed in recent years. The traditional image segmentationalgorithms use a texture or boundary information in a split, while the Grab Cutalgorithm combines the texture and boundary information for image segmentation.Itintegrats a variety knowledge about the global optimal solution and iterativesegmentation. By adding users’ subjective direction, the method has set accuracytarget.Focusing on the problems for2D to3D edge blur and object occlusion in thedaily photo, it is very difficult to use the traditional Grab cut matting algorithm toextract the region. In this paper, the Grab cut is studied and present some goodsolutions to this kind of problems, which comes from an effective solution to the2Dto3D Image Edge Thinning and reducing the complexity of the algorithm point. Andit proposes some advice on the algorithms hoping to coarse to fine segmentationstrategy in future.Firstly, it used the Grab cut to get the initial results on the imagethrough hard partition. Secondly, it reused the Closed-form matting algorithm insteadof the Border Matting algorithm for image edge more blurred and obscured areatransparency to obtain the a value.Only by solving a sparse linear equations,you canobtain the wanted segmentation results.Finally, it uses the2D to3D application to test the improved algorithm.Theresult shows that the improved algorithm could precisely extract the object in lowcomputation time, which has provided a reliable basis for the2D to3D conversiontechnology to the next depth information.