Stroke-Based Chinese Character Completion

In this paper, we propose a method for completing the Chinese character in the image of Inscription rubbings or Chinese calligraphy that is corrupted by fissure and dirt. The method aim at completing the Chinese character from single image based only on the information the image contains. In our approach, for completing the corrupted image with Chinese characters, the structure of the Chinese characters is analyzed firstly using a modified stroke extraction method. All the contours of undamaged Chinese characters in the image are extracted and decomposed into a plausible set of strokes for reducing the problems by the strokes crossing or touching. Then, the structure of the characters in the missing regions is recovered by finding the similarity characters or similarity strokes though a partial contour matching method. Finally, we complete the image use the method with structure propagation, but the curves are not specified by user and are the contour of damaged characters we have recovered.