Fig. 6From: Efficient evaluation of the Number of False Alarm criterionSecond set of results on difficult cracks. These cases push the boundaries of the proposed approach with respect to either sub-crack shortness, or crack thinness or occlusion of crack subparts: original image (1st column), seeds and results of window level detection (2nd column), comparison between ground truth and image level detection (3rd column), fine crack detection after post-processing (4th column)Back to article page