Publication

Abstract : Touching Line extraction from handwritten documents is one of the pertinent problem in the process of optical character recognition, which is generally called as line segmentation of text document images. Identification of the touching or overlapping portions in the document is the major challenge in this regard. In this paper we propose an algorithm for the touching or overlapping line extraction using horizontal projection profiles and a set of predefined chain code sequences that indicates the presence of touching portions between the touching lines. The matched chain code sequence is further normalized to obtain the first difference of the matched chain code sequence. The midpoint of normalized chain code is approximately considered as the segmentation points in the path of line segmentation. The results of proposed algorithm are satisfactory with an accuracy of about 89%. © Research India Publications.