Improvement on Canny operator by algorithm of self-adaptive determining double-threshold

Traditional Canny-operator has a satisfied character in edge detection. But if gradient image is processed with module non-maximum suppression, double-thresholds should be evaluated to detect edges when Canny algorithm is adopted. For different images, the thresholds should not be identical. Currently the double-thresholds are normally obtained through personal experience, which restricts the use of Canny operator in practice. To solve this problem, an algorithm which can adaptively determine the double thresholds based on gradient histogram and minimum interclass variance and improve Canny operator was proposed. Theory and experiments show that the proposed algorithm is effective and correct.