Due to the complexity in structure and the various distortions (translation, rotation, shifting, and deformation) in different writing styles of Handwritten Chinese Characters(HCCs), it is more suitable to use a structural matching algorithm for computer recognition of HCC. Relaxation matching is a powerful technique which can tolerate considerable distortion. However, most relaxation techniques so far developed for Handwritten Chinese Character Recognition (HCCR) are based on a probabilistic relaxation scheme. In this paper, based on local constraint of relaxation labelling and optimization theory, we apply a new relaxation matching technique to handwritten character recognition. From the properties of the compatibility constraints, several rules are devised to guide the design of the compatibility function, which plays an important role in the relaxation process. By parallel use of local contextual information of geometric relaxationship among strokes of two characters, the ambiguity between them can be relaxed iteratively to achieve optimal consistent matching.
展开▼
