Line Segmentation in Persian Handwritten Documents Based on a Novel Projection Histogram Method
Abbas Zohrevand, Javad Sadri, Zahra Imani, Mohammad Reza Yeganezad · 2019
The present paper introduces an efficient algorithm to detect and extract lines in Persian handwritten text images. The proposed approach first calculates the projection profiles (image histogram) of the binary image along with all the rows in the image. Document histograms are noisy functions and there are many spurious local maxima and minima. To simplify measuring the variations in these histograms, a technique based on cubic spline smoothing is applied to smooth the noises. Afterward, by an efficient search of the smoothed histogram, the skew angle (θ) is estimated and the image is de-skewed around the skew angle (θ). Finally, the de-skewed image text is segmented into lines by an efficient linear algorithm with a θ (n). The experimental results in the handwritten Persian database show the superiority of the proposed method in extracting lines in the text images.