Stereo Matching Algorithm Based On Sequence Aligment

Stereo vision is an important part in computer vision, that is an aesthesiatechnology of three-dimensional and distance information. Images taken fromdifferent positions in the same scene, been matched under stereo matchingtechnology, generated disparity image. According to the disparity image, we canreconstruct the three-dimensional information of the scene. In the stereo visionprocessing, stereo matching is one of the key and most complicated process. In thispaper, we study on the efficiency and accuracy of stereo matching algorithm andrealized stereo matching algorithm based on sequence aligment based on biologicaltheory.Firstly, the paper sets the domestic and international situation of the stereomatching, analysises of traditional dynamic programming algorithm and itsshortcomings and point improvement direction. In the same time, it introduces thebasic principles of stereo matching, including key technology, matching constraints,in order to do theoretical support for improving algorithm.Secondly, Sequence aligment algorithm is seriously studied, Three similatitiesare studied in order to testify the feasibility. There are the similarity of gene sequenceand the epipolar, the score function and the disparity map, sequence comparision andthe search strategy of sterep matching. At last, score function is realized.Thirdly, Noticed the poor real-time of matching algorithm and the advantage ofsequence alignment which are parallel and short time. Sequence aligment theoreticalis intodueced into stereo matching algorithm through three step, which are pixel ofeach polar into sequence gene, using branch matrix attained by comparison ofgenomic sequence to establish two-dimensional parallax stereo images and pathsearch optimizing on two-dimensional parallax stereo images.Finally, It realized stereo matching algorithm based on sequence aligment andrealized under VC6.0environment. Through the dense disparity map, it can see thatour algorithm improving the matching speed and greatly reducing false rate.