Chinese Character Recognition Based on Energy Value of the Dual Peripheral Coordinates

Abstract

Feature extraction is the most important part in the process of Chinese character recognition (CCR). When there are such factors as deformity, tilt, uneven illumination existing on the text images, conventional recognition methods such as cellular feature or the linear density characteristics show a degrade on recognition performance. In this paper, we propose a new feature extraction method for CCR based on the energy value of the dual peripheral coordinates. During registration, training samples of the Chinese characters are transformed into binary images. Then the character feature vector is extracted by calculating the square of the coordinates of the peripheral points. During recognition, test samples of Chinese characters are isolated from text images by the steps of image correction, layout analysis, and image segmentation, then the feature vector of dual energy value of the peripheral coordinates is extracted. Character recognition is achieved by searching the minimum Euclidean distance between the testing samples features and the registered training samples features. Experimental results on scanned images and photo images show the effectiveness and accuracy of the feature of dual peripheral coordinate energy value approached in this paper.