Character Recognition Based on Algorithmic of Combining Dynamic Threshold Segmentation and Complementary Similarity Measure

Threshold segmentation and Complementary Similarity Measure plays an important role in the Optical Character Reader (OCR). Generally, threshold segmentation and matching are separated. One of some methods such as iterative method, ENT, OSTU and so on is applied to threshold segmentation. But in practice, these methods are adaptive to a restricted condition. When character is distorted by random noise, these methods of threshold segmentation are not available. In this paper, we will propose a new algorithmic of integrating Dynamic Threshold segmentation and Complementary Similarity Measure to robust above defection. And this algorithmic can use similarity relationship to make the measure robust against noise and improve rate of character recognition greatly.