Segmentation of Touching Modifiers and Consonants in Middle Region of Handwritten Hindi Text

摘要

Character segmentation is a major problem in any text recognition system. The most common problem in the recognition of handwritten text is the presence of touching of left modifier with the consonant in the middle region of the word. In Optical Character Recognition (OCR) system, the recognition rate decreases due to the presence of touching characters. The determination of presence of touching characters is a very tedious task. Based on the structural properties of the text, a new algorithm is proposed to segment left modifier from the consonant in the middle region of the word. The results obtained with the proposed algorithm are very challenging. The problem of over segmentation that occurs in the segmentation of touching left modifier from consonant in the middle region, are also explained.