Firefly Based Word Spotting Technique for Searching Keywords from Cursive Document Images

Abstract

In the fast pace development of digitized technologies, document images have become more fashionable for an information management system present in libraries, organization and educational institutions. Searching information from the document image is very difficult to perform as it compared with digital text. Optical Character Recognition (OCR) is employed to detect the characters and converts the images into their text format. OCR system is not properly converts the various fonts, styles, size, symbols, dark background and poor quality of the document images, however it’s not an efficient method. For this reason, there is a necessity for a searching strategy to find the user specified keywords from document images. Word spotting is an alternative method, whereas keyword is identified without changing the document images. The primary objective of this research work is to search the keywords from printed cursive English document images using word spotting techniques. In this research work, the Firefly based word spotting technique is proposed to search the keyword based on query given by the user. To estimate the efficiency the Firefly technique is compared with existing Enhanced Dynamic Time Warping (EDTW) technique. From the experimental analysis, the proposed Firefly based word spotting technique has produced high accuracy rate and less execution time compared with existing EDTW technique.