Towards Effective Crowd-Assisted Similarity Retrieval of Large Cursive Chinese Calligraphic Character Images

Abstract

Chinese calligraphy is the art of handwriting which draws a lot of attention for its beauty and elegance. People can easily access and enjoy these priceless calligraphy works through the Internet as more and more ancient Chinese calligraphic scripts are digitalized. Despite some research on shape-based retrieval, it is still a great challenge to accurately retrieve the cursive Chinese calligraphy character image(CCI) due to the randomness and complexity of the shape. The paper proposes an effective and efficient crowd-assisted retrieval method of the CCIs which includes three supporting techniques: 1) a NRP- based similarity measure to represent calligraphic character shapes by their contour points extracted from the CCIs; 2) a Hybrid- Distance-Tree(HD-Tree)-based high-dimensional indexing scheme to boost the retrieval performance; and 3) a crowdsourcing-based human verification scheme to refine the result CCIs. Our extensive experiments have demonstrated the satisfactory performance of our proposed retrieval and indexing schemes, respectively.