A new fast and robust template matching with randomness
Chang Liu, Yongqiang Bai · 2017
Template matching is one of the most important techniques in computer vision, where the algorithm should find the location of template image in scene image. The commonly used method of template matching is Normalized Cross Correlation which has a high matching accuracy while consuming a large amount of computational speed. In this paper, a novel, fast and robust template matching approach is proposed. The new algorithm randomly visits the pixels and locates local maxima by gradually moving to the regions with larger NCC values. To further improve the speed and accuracy of the algorithm, several additional rules are established. Theoretical analysis and experimental results show that the proposed algorithm maintain a high matching accuracy while providing a significant speedup.