The Algorithm For The Depth Restortion Of Different Scenes Based On SSVM

The depth information of the objects in the scenes play an important role in many differentfields, such as medical images,3-D scenes reconstruction, remote sensing and so on. Thetraditional methods for gaining the depth information is mainly divided into2classes: the activemethod and passive method. The active method captures the structural information supplied byspecial light source. Then, the passive method mainly based on the thesis of stereo vision. Inorder to avoid the unnecessary computing and measuring, and directly obtain the depthinformation from the scene images, this paper puts forward a new method for the depthrestoration of different scenes based on machine learning method.Recently, Support Vector Machine(SVM) had a full development. Based on the frameworkof SVM, Structural Support Vector Machine(SSVM) involved the data structural can be used forthe structural prediction. So, based on SSVM, this paper constructed a new model for the depthrestoration of different scenes.This paper puts forward a new method for depth restoration based on SSVM, mainlystudied and solved the following problems.1. Construct a new model for depth restoration based on SSVM. According to the imageenergy terms and the framework of SSVM, we formed a relationship between the image featuresand the depth. Then, we constructed a new model for depth restoration based on SSVM.2. Study different image features and apply the optimal algorithm for the optimal solution.In this paper, we studied different features including colors, textures and edges. And in order toapply the optimal algorithm, we formed a constraints based on the sub-modular condition forgraph-cut.3. Improve SSVM by slack rescaling. Due to promoting the accuracy, slack rescaling wasused to improve SSVM. And in order to make it useful, an approximate algorithm was given outfor slack rescaling.