Hierarchical Approximate Matching for Retrieval of Chinese Historical Calligraphy Character

As historical Chinese calligraphy works are beingdigitized, the problem of retrieval becomes a new challenge. But,currently no OCR technique can convert calligraphy characterimages into text, nor can the existing Handwriting Character Recognitionapproach does not work for it. This paper proposes a novel approach toefficiently retrieving Chinese calligraphy characters on the basis ofsimilarity: calligraphy character image is represented by a collectionof discriminative features, and high retrieval speed with reasonableeffectiveness is achieved. First, calligraphy characters that have nopossibility similar to the query are filtered out step by step bycomparing the character complexity, stroke density and strokeprotrusion. Then, similar calligraphy characters are retrieved andranked according to their matching cost produced by approximate shapematch. In order to speed up the retrieval, we employed high dimensional datastructure --- PK-tree. Finally, the efficiency of the algorithm isdemonstrated by a preliminary experiment with 3012 calligraphy characterimages.