Study Of The Handwritten Chinese Character’s Thinning Algorithm

With the advent of the Internet and intelligentized information age, the technologyfor identifying Chinese character has become more mature. However, the identificationof handwritten Chinese character still remains a difficult point of the technology. Due tothe diversity and randomness of handwritten Chinese characters, the identifyingmethods can not be used for all characters. Different methods have to be taken forhandwritten Chinese characters in different styles. The thinning algorithm has alwaysbeen the bottleneck for Chinese characters identification based on structural feature.Thispaper focuses on the thinning algorithm for handwritten Chinese character.This paper studies the preprocessing procedures of the image before thinning,which include graying, binaryzation, line segmentation, characters segmentation,De-noising, etc.The focus is on the existed algorithms’Thinning principle and analyzingtheir merits and demerits. Before thinning the image, we design template to eliminateinfluence caused by concave and sentus. Then We propose an improved algorithm basedon the Hilditch Algorithm by combining it with trend analysis of Strokes for thinning.For the brachylast and unipixel phenomenon appeared after thinning, we continue tomanipulate the pixel on the thinning result until the phenomenon is completelyeliminated. Finally, we set up a certain expansion operator to expand the processedimage and thin it using the method above and get the final thinned resultThe experimenting environment indicated in this paper is VS2008. And thehandwritten characters are taken from HIT-MW Database. The thinning algorithm isprogrammed through Opencv and C language.