Application of RLSA for Skew Detection and Correction in Kannada Text Images

Abstract

The presence of the skew in a captured document image through a photographic camera, mobile camera or scanner is inevitable. In a document image detection and correction of skew are challenging phases before further processing like segmentation and analysis. In this paper, Run Length Smoothing Algorithm (RLSA) is proposed for the detection and correction of skew for handwritten Kannada document images. The proposed work has mainly two parts, the first part is preprocessing of a document using methods like thresholding, the maximum gradient for extraction of text and text line area with no loss of any data. The second part is skew detection and correction. The algorithm RLSA is used row and column-wise of a document image. The RLSA is applied for skew detection to determine skew (slant) angle further the document is turned in the anti-clockwise direction with the preferred angle, which will remove the skew of a document that has occurred while taking the photocopy of the document. The performance proposed method is evaluated for handwritten Kannada documents; the experiment outcomes are significantly better.