Research On Chinese Character Image Recognition

The recognition of Chinese characters is an essential applied field of computer vision and pattern recognition technology. Recently, much as the progress of Chinese characters recognition has been made, there still exist difficult points on how the classify-related image feature is extracted and mutual-information entropy obtained. The accurate extraction and recognition of feature is still the most important things in the recognition of Chinese characters.The following are the major contents in the thesis:1, The extraction of Chinese characters recognition feature. Studies on how the stroke feature information of horizontal, right-diagonal, vertical and left-diagonal is extracted from Chinese characters image two approaches are proposed here: 1) Chinese characters stroke thinning and extraction base on mathematical morphology. 2) Stroke extraction algorithm approach based on Gabor filtering. The structure information of horizontal, right-diagonal, vertical and left-diagonal can be accurately extracted from the original Chinese characters by two approaches.2, The construction of feature model and the code approach for Chinese characters. A novel set of characters feature and code approach is constructed, which is based on the recognizably and extractable of Chinese characters feature, by stroke which is selected as basic unit and the successively structural and statistical extraction approach, including the feature information of stroke which is defined by fuzzy mathematics.3 , Chinese characters recognition algorithm. According to the structure model and code approach, a set of effective recognition algorithm is constructed by the bulb algorithm to subordinately permute the characters feature. At last, a set of Chinese characters recognition system with reaction are constructed, based on and using the error estimation approach of Chinese characters outlines feature.