Skew Detection in Binary Image Documents Based on Image Dilation and Region labeling Approach

In this paper we propose a new and efficient method for estimation of skew angle in a binary document image, based on image dilation and region labeling technique. The input document is dilated by using structuring element as line whose length is fixed experimentally and the region labeling technique is applied using depth first search. Orientation angle is calculated for all the labeled regions, the average of all orientation angles is considered as the skew angle of the document. The experimental results show that better accuracy of estimation could be achieved using this approach since it is based on orientation angles of all the text lines of the underlying document and is the minimum variance unbiased estimator of the true skew angle. The novelty of the proposed method is that, it is robust for machine printed document of any size/font, multi-column layouts and documents containing graphics, pictures, charts, tables etc