The utilization of closing algorithm and heuristic information for broken character segmentation

In Thai printed character recognition systems, an important problem that decreases the accuracy is broken characters. These characters could cause the error in segmentation process. To solve this problem, a method for broken character segmentation in Thai printed document is presented. It consists of two main steps: text line detection, for extracting text lines from an image, and character segmentation, for extracting broken characters from a text line. The character segmentation consists of four steps: Gap reduction using closing algorithm, character segmentation using space, large character splitting and small character merging using heuristic information. The advantage of this approach is the ability to segment broken character even when it is split into a large number of segments. The experimental result shown that our method achieves 91.09%