Implementation and improvement of area-based stereo matching algorithm

In order to obtain the relationship of matching cellsbetween two homologous images accurately and to solve the problem of large computation in the area-based stereo matching method, an area-based stereo matching method is presented in the paper. Based on rectifying the images by searching range from two dimensions to one dimension, an implicit constraint algorithm and an SMP algorithm with a uniqueness constraint is combined to achieve the matching of stereo pairs. Then, several fake matching points in the primary matching process are detected and removed by using the false epipolar constraint method. Finally, a disparity map is interpolated. Experimental results show that the accuaracy of the presented method is improved by 10% as compared with that of conventional SMP algorithm. In addition, a Box-filtering accelerating method is also implemented to decrease computational complexity, the program running time is about 50% that of BM algorithm. These results show that the improved algorithm obtains higher accuracy and satisfies the requirement for a real-time image processing system.