Segmentation of Handwritten Text Using Bacteria Foraging Optimization

Text collected for digitization in the form of handwritten paragraphs is difficult to segment into individual characters. Identification of whole words is more difficult, so the idea is to segment the word into individual characters. These characters will be joined to make artificial words. Actual words may be predicted from these artificial words by any word prediction algorithm. A novel bacteria foraging optimization-based pixel model has been used to optimally segment all valid characters from the word. The bacteria colonies arrange in a specific shape determining a character. The offspring bacterial colonies produce the candidate segmented characters. The unhealthy colonies from the candidate segmented characters are eliminated deducing the pruned offspring bacterial colonies identified as optimal segmented characters. Paragraphs written by more than 50 subjects have been segmented with higher character segmentation accuracy.