Theory And Research Based On Energy Minization For The Graph Cuts Stereo Matching Of Algorithm

Stereo matching is an important field in computer vision, which can obtain thedisparity map by matching two or more images so that we can calculate the3Dinformation of objects in the scene. At present, the stereo matching technique has beenwidely used in the field of3D reconstruction, military navigation and virtual reality. Basedon the principle of stereo matching algorithm, this paper focuses on the graph cuts ofstereo matching algorithm on energy minimization, the main research and work are asfollows:First, it briefly introduces the principle of stereo matching algorithm andclassification, and analyses the research emphasis and development trend of stereomatching algorithm in detail.Secondly, this paper proves the relationship between energy minimization and thegrid map in the process according to the theory of graph cuts of stereo matchingalgorithm,, and how to achieve the maxflow/mincut algorithm.Finally, the α expansion algorithm、α—βswap algorithm are introduced in this paper, and theoptimizations and improvements in these algorithms are proposed. According to the α expansionalgorithm which is proposed by Boykov, to optimize flow structure in the algorithm. The improvedalgorithm firstly introduces a new mechanism to stop the cycle when the structure parameters meet thestop condition and the matching process is terminate, at the same time, to optimize the “α” randomstate in the process of expansion moving. The simulation results shows that the improved algorithmachieves an accurate disparity map which is similar to the original algorithm, in addition, it savesmuch time in computation...