Fig. 11
From: Efficient cost aggregation for feature-vector-based wide-baseline stereo matching
Representative examples of disparity estimation from the 2014 Middlebury Stereo Dataset. Columns 1 and 2: the stereo pairs. Column 3: disparity estimation results by the proposed method, where pink denote pixels with incorrect disparity estimations. Column 4: comparison of the proposed method with Min et al.’s method [14]. If the disparity of a pixel is correctly estimated by the proposed method but incorrectly by Min et al.’s method, the pixel is shown in green color; otherwise, it is shown in red color. Column 5: comparison of the proposed method with the census transform using the same color convention. A blue border around the image indicates the case where the proposed method performs worse compared to the other methods