Abstract: This paper presents an stereo matching algorithm based on an image segmentation and Ground Control Points （GCP）. Regions of homogeneous color are located via the Mean-shift color image segmentation method. The combination of RGB information of the pixel and the gradient information is used to calculate the initial disparity. Under the constraint of Ground Control Points （GCP）, the energy function is constructed, and the minimum energy function is computed by using Dynamic Programming （DP） method. The initial disparity is optimized by voting and the final disparity map is acquired in the area of image segmentation. The experimental results show that the algorithm has a better performance in occlusion region and disparity discontinuous region, and it is worth mentioning that the problem of stripes caused by DP algorithm is also solved.

Key words: stereo matching, image segmentation, ground control points, Dynamic Programming（DP）, disparity

摘要： 提出了一种基于图像分割和地面控制点（GCP）的立体匹配算法。利用Mean-shift算法将参考图像根据彩色信息快速聚类成不同区域，利用像素点的RGB信息与梯度信息相结合计算初始视差;引入地面控制点（GCP）约束，构造能量函数，利用动态规划方法（DP）计算能量函数最小值；在图像分割区域内采用快速投票方式优化初始视差并获得最终视差图。实验结果表明：该算法能有效处理视差不连续和遮挡区域，也解决了DP算法带来的条纹等问题。

关键词: 立体匹配, 图像分割, 地面控制点, 动态规划（DP）, 视差