Optimized Word Segmentation for the Word Based Cursive Handwriting Recognition

This paper is the result of research carried out to enhance the efficiency of cursive handwriting word based segmentation for sigma based offline cursive handwriting recognition. After success in recognition process the optimization process was undertaken. A new algorithm with the combination of smart data structure techniques was successfully developed and tested over various samples. Also the comparative analysis was taken in extensive research between bitmap and bitmap data. The algorithm was tested on both type of images and results under different circumstances were compared. Each image type had its advantaged and disadvantages. The segmentation time was well reduced by using basic iteration for just reading the image else every calculation and word segmentation was handled by different data structures depending upon their use and function they handle the best in different situations. Binary image made it faster with loss of some quality while with retaining the quality the speed was still better than before. The idea of using during-process image scaling and re-scaling is also under consideration and experiment. The input images to the algorithm were already processed, normalized and noise removed by edge detection techniques using very careful threshold to have the difference between a word and a noise. This segmentation algorithm is one of its kinds as it has been for the first time applied on word based handwriting. During the comparison search with other algorithms only one paper was found which used almost the same technique but it was for character based handwriting recognition [13].