Language Independent Skew Detection and Correction of Printed Text Document Images: A Non-rotational Approach

摘要

The work presented in this paper focuses on detection and correction of skew in language independent printed text document images, ranging from -30° to +30°. A method is developed to detect the skew in printed document image and a non-rotational transformation process is proposed to correct the skew. The method uses linear approximation approach directly to correct the skew without finding the angle of skew. The method consists of two stages. The direction of skew is detected based on fixing boundary lines across the corners of the document image in the first stage. Second stage uses a non rotational transformation to correct the skew in the document image. The method has been tested on variety of printed document images with different skew angles and languages.