A new algorithm for skew correction and baseline detection based on the randomized Hough Transform

Abstract

The proposed technique is based on the detection of the lower baselines of the text lines of Arabic documents. As the lower baseline pixels belong to the lower edge of the word images, we first locate vertically the black–white transitions at the black pixels where the resulting image would emphasize the baselines of the text. Once the skew angle is determined using a randomized Hough transform, the baselines are extracted using y-intercept histogram. This algorithm can also contribute significantly for text line extraction from skewed document images for many languages.