Calligraphy Writing Style Recognition Technology

Chinese calligraphy is a valuable part of the Chinese culture heritage. With the continuous development of digitization technology, more and more calligraphy works are digitized, preserved and exhibited in digital library. Users always want to appreciate the style-similar works simultaneously, but the recognition technique is not mature. Some users even can’t distinguish the writing style of calligraphy works. So automatic recognition method based on computer vision is needed.In this paper, we proposed a global descriptor based SSC(Similarity Sensitive Coding) method and a local descriptor based feature library filtering method. The first method use GIST feature to represent character images, by converting the GIST vectors to binary vectors using SSC algorithm, the style similar images become neighbors under hamming distance. The second method use SIFT feature to represent the character images, after applying a feature filtering procedure, the style-irrelevant SIFT points can be wiped out and a feature library is established to do calligraphy writing style recognition. Experiment shows that both of the methods have advantages and disadvantages. The SSC based method is robust and fast, but the feature library filtering based method is more accurate when applied to printed characters.Further, we take advantage of that the SSC based method encodes the feature vectors to binary codes which makes feature codes smaller and matches faster, build a Chinese character recognition system on Android. We applied SSC and GIST extraction on Android and used JNI to implement the core algorithms, developed a Chinese character recognition software, which can identify simplified and traditional Chinese characters.