Near-real-time stereo matching method using both cross-based support regions in stereo views

摘要

This paper presents a near-real-time stereo matching method using both cross-based support regions in stereo views. By applying the logical AND operator to the cross-based support region in the reference image and target image, we can obtain an intersection support region, which is used as an adaptive matching window. The proposed method aggregates absolute difference estimates in the intersection support region, which are combined with the census transform results. The census transform with a fixed window size and shape is applied, and only the resultant binary code of the pixel in the intersection support region is used. From Middlebury images and their ground truth disparity maps, we compute the area similarity ratio of support regions in stereo views. Then, a conditional probability of observing a correct disparity estimate with respect to the area similarity ratio is examined. By taking a natural logarithm of the probability, a relative reliability weight about the area similarity of support regions is obtained. The initial matching cost is then combined with the reliability weight to obtain the final cost, and the disparity with the minimum cost is chosen as the final disparity estimate. Experimental results demonstrate that the proposed method can estimate accurate disparity maps.