Rectification of Camera Captured Document Images using Component Analysis

Abstract

Image distortion is a common problem when one captures a document page by a digital camera or mobiles from a certain angle. It results a horizontal and vertical distortion of the text lines in the digitized document. This type of distortions will not only impair the readability, but also reduce the accuracy of any Optical Character Recognition system. In this paper we propose a novel technique to identify the entire text-region within the document page and then we apply the homography transformation on the corner points of estimated test-region for rectifying its distortion. Compared to other existing methods, the present system is computationally less expensive. The experimental result shows that, our method is able to correct any type of distortion effectively and outperforms the state-of-the-art methods.