A Rotated Character Recognition Method Based on Geometry Correction

In the field of character recognition, the problem caused by geometric deformation or the rotation of characters leads to a low character recognition rate. In this paper, a high speed recognition method based on affine invariant to recognize the deformed characters is proposed. In this method, the deformed character is rectified within a circle using the moments of deformed character, and then it can be recognized in real time by comparing the similarity of distance histogram and angle histogram. Experimental results show that the method based on affine invariant can recognize the English characters and numbers with a very high recognition rate. On the other hand, this method can recognize the rotation characters by a high speed comparing with the traditional method.