Analysis of Different Techniques for Segmentation of Connected Handwritten Indian Scripts Words

In OCR processing text is extracted from an image. Here many common imperfections occur, like the characters that are connected together are returned as a single sub-image containing both characters. So this is the major challenge in the recognition process and we need to segment the connected words to get correct results. In segmentation process the first task is to find out the segmentation point. Once this is done the next step is to fragment the image from that point and get two separate sub-images for each character. There are many algorithms proposed for this process. One class of approaches uses the straight segmentation technique and segments the words based on the vertical and horizontal histogram profiles. Another approach uses contour features of the component for segmentation. Some researchers segments the characters by recognizing the profile features of touching character and the segmenting them. This technique is commonly known as recognition based segmentation or cut classification.