A Eeature-Constrained Stereo Matching Algorithm
一种基于特征约束的立体匹配算法

In the situation of image synthesis for image based rendering(IBR), we need not only precise matching but also a dense disparity map. In order to meet the needs of this application, a feature constrained stereo matching algorithm is presented. It integrates feature based and area based process. First we obtain a sparse precise disparity map of edge points using feature based approaches, next we fill in the disparity data of featureless points by using area based algorithm under the constraint of the acquired disparity map. In this way, not only can we reduce the search for correspondences but also ensure the validity of matching. During the matching phase the two images play a symmetric role to ensure the validity of matching. At the same time, the technique of stereo matching with adaptive window based on the gray value information of nearby points is adopted in the area based process. The experimental results show this algorithm is effective and it is of great value to use.