基于互相关信息的非参数变换立体匹配算法

针对传统的非参数变换立体匹配算法实时性不强和可靠性不高的局限性,提出一种基于互相关信息的非参数变换立体匹配算法.将变换窗口内所有像素的灰度值进行平均,然后将平均值作为中心像素的灰度值;为了在立体匹配时能够考虑像素间的互相关信息,将变换窗口各邻域与中心像素的相对位置大于一个单位的像素,其灰度值用双线性插值后的灰度值替代;将变换窗口的像素灰度值进行非参数变换立体匹配,得到致密的视差图.实验结果表明:与其他基于局部的单一立体匹配算法相比,该算法得到的误匹配像素百分比与其他算法相当,能够有效提高传统非参数变换立体匹配算法的鲁棒性.

Mutual information based non-parametric transform stereo matching algorithm

Aiming at the limitations of the weak real-time and the low reliability in the traditional non-parametric transform stereo matching algorithms, a mutual information based non-parametric transform stereo matching algorithm was proposed. Gray values of all the pixels in the transform window were averaged, and then the mean value was taken as the gray value of the center pixel. In order to take the pixels' mutual information into consideration while finding stereo correspondence, the original gray value of the neighborhood pixels whose the relative position was one unit greater than that of the center pixel was replaced by the gray value through bilinear interpolation. The non-parametric transform of the gray values of the pixels in the transform window was implemented, and a dense map was obtained. The experimental results indicate that compared with other single area-based matching algorithms, the percentage of bad matching pixels is nearly equivalent to other algorithms while it can improve the robustness of the traditional non-parametric transform stereo matching approaches effectively.