An Efficient Character Segmentation Algorithm for Recognition of Arabic Handwritten Script

摘要

This manuscript proposed a new novel algorithm for vertical segmentation because the accuracy of recognition will be very high if there is a good character segmentation. For the purpose of locate the segmentation points used the word image thinning to fetch the width of stroke of one pixel and to detect the ligatures of Arabic characters geometry and shape are used in the segmentation procedure. The proposed segmentation approach is worked with touching characters, in case of ligatures touching segmentation existing between the characters of consecutive closed and the existence of ligatures within characters in the case of open characters. It evaluated on IFN/ENIT, AHDB and our dataset. It shows higher accuracy with an excellent performance by decreasing over-segmentation problem which showed through open characters segmentation and correct the segmentation in cases of the touching characters or the miss-segmentation errors in images of word.