Segmentation of Handwritten Documents Containing Kannada Script

Segmentation is one of the important phases of Optical Character Recognition (OCR) system, which extracts objects of interest from an image. Feature extraction and classification phases of OCR will be more effective, if the techniques selected for segmentation is effective. This paper focuses on to develop a system for handwritten documents containing Kannada script and proposes suitable techniques to perform preprocessing and also segmentation such as line, word and character segmentation. Novelty is achieved by proposing a modified horizontal projection profile method for line segmentation, in which well separated lines and overlapping lines are detected. An average accuracy of 97.5% is achieved for line segmentation and word segmentation.