Character Retrieval Based on the Improved Contour Feature and Texture Feature

In order to retrieve the character images but avoid the OCR system which is limited in some application, a method for Chinese characters retrieval is presented in this paper. It firstly extracts the improved contour feature of Chinese character images, which is used to match roughly. Then, the candidate character images attained above are divided by the centroid of each one. For each part divided, the texture feature of the Fourier spectrum is extracted for scrutinized detection to obtain the final results. The experiment results imply that the method proposed is efficient in retrieving different grays, different fonts and sizes of Chinese character images.