A novel non-parametric transform stereo matching method based on mutual relationship
A novel non-parametric transform stereo matching method based on mutual relationship
复制标题
一种基于相互关系的新型非参数变换立体匹配方法
DOI：
10.1007/s00607-018-00691-3
复制
发表时间：
2019-06
期刊：
影响因子：
3.7
通讯作者：
Huang Peng
中科院分区：
文献类型：
--
作者：
Lai Xiaobo;Xu Xiaomei;Lv Lili;Huang Zihe;Zhang Jinyan;Huang Peng
针对目前绝大多数局部立体匹配方法高度依赖图像强度统计特性的问题，提出了一种基于相互关系的非参数变换立体匹配方法。对传统的非参数变换进行了研究，分析了其局限性。为了在寻找立体对应时考虑像素的特殊位置信息，将相对位置比中心像素大一个单位的邻域像素的原始灰度值替换为其周围四个像素的灰度值内插。然后进行新的非参数变换立体匹配。用标准图像数据集和真实场景采集的图像对该方法进行了测试。实验结果与基于灰度的算法进行了比较，匹配不良像素的百分比与其他检测算法基本相当，并且在实际条件下表现出了较好的性能。
To cope with the problem of the vast majority local stereo matching approaches that rely highly on the statistical characteristics of the image intensity, a novel non-parametric transform stereo matching method based on mutual relationship is proposed. The traditional non-parametric transform is investigated, and its limitations are analyzed. In order to take the pixels’ special location information into consideration during finding stereo correspondences, the original gray values of the neighborhood pixels whose relative position is one unit greater than that of the center pixel are replaced by the gray values interpolation of the four pixels surrounding it. Then the new non-parametric transform stereo matching is performed. The proposed approach is tested with both the standard image datasets and the images captured from realistic scenery. Experimental results are compared to those of intensity-based algorithms; the percentage of bad matching pixels is almost equivalent to the other examined algorithms, and the proposed algorithm exhibits robust behavior in realistic conditions.