Corner Detection of the Chinese Characters Based on First-Order Difference

In view of the importance of the Chinese characterspsila corners, as well as the complexity of Chinese characters themselves, an algorithm designing for detecting Chinese characterspsila corners based on first-order difference is proposed in this paper. According to smoothing the eight-neighborhood chain codes of the Chinese characterspsila contour, the proposed algorithm can avoid the interference of the saw tooth phenomenon in Chinese characters and some spurious corners efficiently; using First-Order Difference of the eight-neighborhood chain codes to represent the rate of the directionpsilas changing of Chinese characterspsila contour makes the computation simple and convenient; threshold proposed in this paper can adaptively change as the length of each different closed boundary changing. To validate the algorithm, tests comparing the proposed algorithm to four corner detection algorithms are given. The results show that the proposed algorithm is not only simple and efficient, but also accurate and stable, moreover it wonpsilat be influenced by the affine transformation.