A Fast and Novel Skew Estimation Approach using Radon Transform

Abstract

In this paper, an effective and reliable skew estimation technique for machine printed documents and photos using radon transform is proposed and is compared with other methods used for skew estimation such as Fast Fourier Transform (FFT), Hough Transform (HT), combination of Horizontal Projection Profile (HPP) and Hough Transform, combination of Gabor filter and Radon transform, combination of Wavelet Transform (WT) and Hough Transform and Horizontal Projection Profile only. The radon transform based skew estimation approach gives faster and better results compared to other methods. The proposed technique is tested on 150 document images with skew varying from 1 to 25 degrees. It provides 100% accuracy for skew estimation with average time of execution 2.23 seconds for scanned document images containing text, 0.1404 seconds for machine printed images containing text, 1.2 seconds for scanned document images containing pictures and 0.0156 seconds for machine printed document image containing pictures.