Skip to Main Content
A method for skew detection of document images is presented. The method is based on the information of the directions of text lines in the images. To extract the directions of text lines in the images, a measure of line-likeness of texts is adopted, which reflects the strength and direction of a pixel being part of a text line. This measure is applicable to texts of various fonts, sizes, and even languages. The skew angle of the whole image is the consensus among all pixels in the image. Several methods for the consensus are proposed. Experimental results of the method on document images are presented.