A Novel Segmentation Algorithm For Handwritten Document Using Contour Point Extraction

In the present world, the financial system, postal service, and insurance firms all write a lot of data on paper. The Optical Character Recognition system is used to recognize the text from these physical paper documents. The most important phase in any OCR system is the segmentation process. The proposed system comes under the area of Image processing. The domain where segmentation for handwritten documents is used is Pattern recognition. One of the main issues with handwriting recognition has been the division of words into individual letters. This project aims to develop an automatic segmentation algorithm to segment characters from handwritten documents and address all the challenges. To achieve better results, different preprocessing techniques and a histogram projection method are used. A Vertical histogram projection is proposed to segment the words from the input image. The lines are segmented by extracting the row-wise existence of consecutive black pixels. The IAM dataset was used to evaluate the effectiveness of the suggested technique. In the proposed method, the accuracy of line, word, and character level segmentation is 99%, 95%, and 85% respectively. A Contour Point Extraction method is used for character segmentation which gives higher accuracy than the existing methods. The segmented characters can be further used to train a Machine Learning algorithm for character recognition.