Two-stage skew correction of handwritten Bangla document images

Abstract

Skewness in the handwritten document images is a common scenario. Therefore, it is very much required to detect and correct the skewness before the document is presented to the document image analysis system. In this regard, the present work develops a two-stage Hough transform based approach to remove the skewness in the document images written in Bangla script. Firstly, page-level skewness is removed by rotating the skewed text lines appropriately and then skewed words in each text line, if any, are also rotated along a reference line.