Edge detection for Chinese text image based on novel differential operator

Consisted with the fast development of digital information technique, the historical paper-base documents haven been planned and executed to be scanned and digitalized. Transformation of the detected character signal into binary codes is essentially decided by the edge detection procedure. Though there have some reports about English alphabet detection, few focused on the Chinese text. In this paper, the differential operator method is put forward to detect the image edge based on the step change between strokes and the grey background. After analyzing the traditional operators, which is effective to detect simple scenery image but deteriorating the Chinese text, novel 5×5 dimensional differential operator is proposed. Results verify that the novel operator is more effective than the custom ones in edge restoration of Chinese stoke. Also, it presents that the denoise procedure is necessary to improve the pixel weighted average performance of the differential operator. The novel operator can be integrated in Optical Character Recognition (OCR) system as preprocessing element.