The input document images with skew can be a serious problem in the optical character recognition system. A method is proposed for skew detection in binary document images using mathematic morphology. The basic process of our approach consists of three steps: Firstly, a dilation operation is applied to the binary image; Secondly, the dilated image is thinned; Finally, the skew angle is detected using the Hough transform. The proposed approach with high precision can detect skew with large angle (-90°|90°). The experimental result shows this method is applicable and efficient.
展开▼