A Segmentation Algorithm for Touching Character Based on the Invariant Moments and Profile Feature

Unsuitable segmentation is the primary cause of recognition errors in optical character recognition. In this paper, after analyzing several representative related approaches, a segmentation algorithm for touching characters based on Hu´s invariant moments and profile feature is presented. Hu´s invariant moments are extended. By using the concave and convex of string profiles, the cut-off points of the candidates of the touching characters are located. And with the width constraining, the best cut-off points using the extended invariant moments are found out. Experimental results show that this approach is effective and feasible, and also reduce the recognition error because of segmentation errors.