Offline touched cursive script segmentation based on pixel intensity analysis: Character segmentation based on pixel intensity analysis

Offline cursive character segmentation refers to extracting individual characters from cursive handwriting on scanned documents. The character recognition accuracy is heavily based on accurate character segmentation. In this process, touched cursive characters segmentation is the main bottleneck towards accuracy rate. This paper presents an efficient technique for offline cursive touched character segmentation based on pixel intensity analysis. Initially, scanned cursive handwriting samples containing touched characters are collected and pre-processed. To avoid processing of the entire image, a region of interest is detected in each collected sample. The region of interest consists of core-region and touched pattern. Finally, based on pixel intensity analysis in the region of interest, a non-linear segmentation path is traced to segment touched cursive pattern into individual characters. The proposed touched cursive character segmentation approach is fully automatic and achieved promising results as compared to other approaches reported in the state of art in most cases.