Grayscale template-matching invariant to rotation, scale, translation, brightness and contrast

摘要: In this paper, we consider the grayscale template-matching problem, invariant to rotation, scale, translation, brightness and contrast, without previous operations that discard information, like detection of edges, interest points or segmentation/binarization images. The obvious "brute force" solution performs a series conventional template matchings between image analyze query shape rotated by every angle, translated position scaled factor (within some specified range scale factors). Clearly, takes too long thus is not practical. We propose technique substantially accelerates searching, while obtaining same result as original brute force algorithm. experiments, our algorithm was 400 times faster than Our consists three cascaded filters. These filters successively exclude pixels have no chance matching from further processing.