Probabilistic and interactive retrieval of Chinese calligraphic character images based on multiple features

Abstract

This paper proposes an efficient probabilistic indexing scheme called Probabilistic Multiple-Feature-Tree(PMF-Tree) to facilitate an interactive retrieval of Chinese calligraphic manuscript images based on multiple features such as contour points, character styles and types of character. Different from conventional character retrieval and indexing methods [18] which only adopts shape similarity as a query metric, our proposed indexing algorithm allows user to choose the above three kinds of features they prefer to as query elements. Moreover, a probabilistic modal is introduced to refine the retrieval result. Comprehensive experiments are conducted to testify the effectiveness and efficiency of our proposed retrieval and indexing methods respectively. © 2011 Springer-Verlag.