A Stereo Matching Algorithm Based on Left and Right Views
Zhao Liang-liang · Jisuanji fangzhen · 2010
An algorithm that improves the matched precision of occlude and textureless region is presented.By using the global constraint conditions,the uniqueness and the order constraint,high credible parallax is obtained from different regions under different constraint condition and intensities so as to solve matching of left and right views from the complicated scenes.Under the constraint of the gradient of parallax,the improved cooperative algorithm is applied to pervade the of the point which has high reliability so that the dense parallax map can be achieved at last.The experiment results indicate that the proposed method can generate accurate dense parallax map to avoid problems of area-matching.