Aug. 26, 2004 to Aug. 26, 2004
Li Chen , Tsinghua Univ.
Xiaoqing Ding , Tsinghua Univ.
Character image contains not only the character information, but also the type information. The character's type information denotes the various category of the character, such as the font families, handwritten or printed, to what language the character belongs, etc. Most of the existent methods for character type recognition are carried out on a group of characters belonging to the same category, and all these methods concentrate on their own specific field. In this paper, a novel universal method for character type recognition is proposed, which is based on a single unknown character and can be used in various fields. We employ a wavelet transform on the character image and extract wavelet features from the transformed image, which are used by a MQDF classifier. Compared with existent methods, our method is much more flexible, robust and effective.
Li Chen, Xiaoqing Ding, "A Universal Method for Single Character Type Recognition", ICPR, 2004, Proceedings of the 17th International Conference on Pattern Recognition, Proceedings of the 17th International Conference on Pattern Recognition 2004, pp. 413-416, doi:10.1109/ICPR.2004.1334141