基于多尺度的小基高比立体匹配方法

针对小基高比立体匹配中的＂黏合＂现象和深度精度问题,提出一种小基高比立体匹配方法.该方法通过将自适应窗口技术和多窗口策略相结合为参考图像确定匹配窗口;然后根据规范化互相关函数和＂胜者全取＂策略计算整数级视差;再以整数级视差为基础利用基于二分法的亚像素匹配方法计算亚像级视差;最后采用基于图像分割的迭代传播方法以获得稠密视差图.实验结果表明：该立体匹配算法减少了小基高比匹配中的＂黏合＂现象,同时获得了稠密的高精度亚像素级视差,其亚像素精度可优于1/20个像元.

Multiscale-based stereo matching method for small baselines

A stereo matching method for small baseline was proposed to solve adhesion phenomenon and depth accuracy in small baseline stereovision.The adaptive window was combined with multi-window to determine the supporting window for the reference image,and then the normalized cross correlation function and the"winner-take all"strategy were used to compute integral disparities.The subpixel stereo matching method based on binary search was employed to obtain sub-pixel disparities on the basis of integral disparities.Finally,the iterative diffusion method based on image segmentation was utilized to attain a dense disparity map.Experimental results show that the proposed stereo matching method largely decreases the adhesion phenomenon and achieves the dense disparity map with sub-pixel precision better than 1/20.