Ancient Chinese Character Image Retrieval Based On Symmetrical Regions

Due to the characteristics of ancient Chinese characters such as complex structures andvariant shapes, there exist many theoretical and technical problems to be solved in featureextracting, clustering and retrieval of their images. In this paper, a retrieval method of theimage with single ancient Chinese character was established with the strategy of “clusteringbefore matching” by integrating the structural features of these character images, which canbe helpful to the experts’ research work on Chinese characters.In the course of ancient Chinese character image segmentation, the images with singlecharacter can be obtained from the process in which the full page of ancient Chinesecharacters is syncopated, deleted and merged through the searching for connected areas. Inthe course of ancient Chinese character image retrieval, the strategy of “clustering beforematching” by integrating the structural features of these character images is adopted; in thestage of clustering, K-means algorithm is employed to cluster all the single character imagesin global and local areas, which is on the basis of the character images already preprocessedbeing divided by the flexible grids being extracted; in the stage of retrieval, the globalfeatures of the Chinese character images to be retrieved are gathered to determine the classesthey belong to and retrieve in the class in the aspect of the global retrieval; in the aspect oflocal retrieval, according to.the retrieval area selected by the users and the symmetricallocation area computed, the original clustering center and symmetrical clustering center areachieved through the area feature matching. The images which has the similar features withthe area selected by users are searched in the corresponding classes and the correspondingretrieval results are shown by controlling the degree of similarity. The experiments’ resultsshow that this method above can satisfy the retrieval need of the researchers on Chinesecharacters, which results in higher work efficiency in the study ancient Chinese characters.