Handwritten Chinese Character Recognition

Off-line Chinese character recognition is an important area in the research of pattern recognition; Chinese information processing is an important interface technology. After decades of research, printed font in the identification, bank checks, post system, these areas have made important achievements. However, throughout the field of OCR, especially in the field of freedom handwritten, there are still considerable difficulties about raising recognition rates in this area, and it becomes the most challenging issue.This paper's main work is as follows:First, this paper does a lot of image's pre-processing which can removes the information that has nothing to do with the text, such as color information. For the general text of the image, the paper designs a method to obtain the prospects of text image. From the effect of the results, on the one hand, this method removes a lot of background information, on the other hand, retains the text information.Second, summarize the characteristics of single character recognition in the handwritten Chinese characters, such as the feature of the outline characteristics, the feature of the direction of the line-stroke, the feature of the grid, and the background feature. This paper detail introduces the method about the extraction of characteristics of the direction of the line based on statistics, as well as the characteristics of the grid research, and analyzes the strengths and weaknesses about these characteristics in Chinese character recognition. Then discuss the classification about thinning algorithm, like they can be divided into pixel-based algorithm and the marginal and erosion -based algorithm. This paper designs a new parallel thinning algorithm based on retaining skeleton points which have good results.Third, studying the characteristics of the multi-line handwritten is this paper' main achievements. First, put forward the labeling method of the branches of the connectivity. Because the method is to deal with the large amount of data, the article first pretreats on the text of the image to segment line to reduce the amount of data. Due to the complex structure of the text components, design a method based on connectivity branch of the merger and the decomposition to get a better effect.Fourth, for Chinese handwriting's prevalence of adhesion, overlapping, this paper summarizes the previous achievement, such as corner detection algorithm, a segmentation algorithm based on the background of the character of the image. And from another angle, that is, the whole text looks as a graph, the segmentation of the image is equivalent to a split from a graph to sub graph, which can help to realize the segmentation about the adhesion character.