Estimating the Skew Angle of Scanned Document through Background Area Information

Abstract

The skew correction of scanned document images is an important step toward an automatic recognition system. Several techniques have been developed to estimate the skew angle of a scanned document. However, most of these techniques have the problem of expensive computational costs. A variation of a fast method based on parallelograms covering is presented in this article. The objective is to obtain lower computation time and provide a way to work well over noisy images and also over different layouts containing non-text areas with no decrease in performance. Experimental study with different databases achieved results that overcome previous techniques.