自适应窗口和半全局立体匹配算法研究

针对区域立体匹配算法中匹配窗口的选择和在图像中视差不连续、弱纹理区域误匹配率较高的难题,提出了一种自适应窗口和半全局算法相结合的立体匹配算法.该算法先通过基于灰度的自适应窗口算法计算初始匹配代价,然后利用相邻像素之间的视差梯度作为约束,采用半全局立体匹配算法以得出视差图.最后对左右视差图进行左右一致性遮挡检测,获得精确的稠密视差图.针对不同的图像信息采用自适应窗口和半全局算法相结合的立体匹配算法,剔除、修正了视差不连续、弱纹理区域的误匹配点,得到匹配效果较好的视差图.实验结果表明:该算法有效降低了视差图在视差不连续、弱纹理区域和遮挡处的误匹配率.

Adaptive Windows and Semi-global Stereo Matching Algorithm

There are still some common issues existing in the region stereo matching algorithm for the selection of matching windows and region matching in disparity discontinuity and weak texture.In order to solve these problems,this paper proposes a stereo matching algorithm based on the combination of adaptive window and semi global algorithm.Firstly,the initial matching cost are calculated based on adaptive window matching algorithm.Then,Semi-global stereo matching algorithm is used,in which there is a constraint by disparity gradients between adjacent pixels,to obtain the initial disparity map.Finally,left-right consistency check is employed to eliminate the errors taken place in occluded regions,and obtained the dense disparity map.The authors take adaptive windows and semi-global stereo matching algorithm for different images to correct mismatching points in the disparity discontinuities and weak texture regions.Experimental results show that the proposed algorithm can reduce the error matching rate effectively,and improve the matching accuracy in the disparity discontinuities,weak texture and occlusion region.