Recognition of Chinese character in gray image based on improved radiant-projection-transform

An algorithm called radiant-projection-transform (RPT) is proposed, which can be used in the recognition of Chinese character in gray image. The mathematical model and implementation method are given. In order to improve the performance of the original projection function, improvement and optimization are applied to the original algorithm. Consequently, a new function called improved-projection-accumulative-function (IPAF) is obtained. The characteristics of IPAF as well as Fourier descriptors (FDs) based on it are studied in the case of gray Chinese character recognition. Based on theoretical analysis and experiments, FDs are utilized as features for recognition. The recognition accuracy of 3755 most frequently used Chinese characters (gray) reaches 98.9 percent by using FDs based on IPAF.