AN ENHANCEMENT OF TEXT DETECTION ALGORITHM WITH THE USE OF STROKE WIDTH TRANSFORM / Lovelle Eriel R. Friginal and Shiela Marie A. Lopez. 6

ABSTRACT: Optical character recognition (OCR) is the electronic conversion of images with printed text into editable text. It is widely used as a form of data entry from printed paper data records, whether passport documents, invoices, bank statement, receipts, business card, mail, or other documents. It is a common method of digitizing printed texts from images so that it can be electronically edited, searched, stored more compactly, displayed on-line, and used in machine processes such as word translation and search, data extraction and text mining . OCR is a field of research in pattern recognition, artificial intelligence and computer vision. The researchers study one of the latest published OCR algorithm entitled Text Detection Algorithm with the use of Stroke Width Transform and found some weak points on the said algorithm. The proponents improved the Text Detection algorithm that is embedded in an optical character recognition application. The study added some enhancements in address to the three major drawbacks of the existing algorithm namely: Ineffectivity of the recognition of concave curve texts; Inaccuracy of recognition of similar looking characters; and Production of high false positive or detected image region where there is no existence of text. The proponents proposed a novel text detection approach that gave solution to the said text detection problems. The project aimed the reduction of the rate of false positives through an enhanced version of the Stroke Width Transform algorithm for the accurate detection and localization of text regions in an image. The proposed algorithm also improved the detection and localization of text regions in an image. The proposed algorithm also improved the detection and recognition of similar looking characters and concave curve texts. The researchers used the Descriptive Research method and Quota Sampling Technique that helped them gather essential information about the text detection algorithm. Through conducting a survey and with adequate interpretation of results, the proponents were able to collect data and identify the advantages and disadvantages of the existing algorithm and used it as their basis in defining the statement of the problems and objectives of the study. The proponents modified the existing algorithm to improve its performance. The proponents provided the necessary information about the existing and enhanced algorithm. They also provided simulations and sample screen shots for the readers to further understand what text detection algorithm is all about. The proponents created an enhanced application that answers the problems stated on this study. An efficient algorithm was used to build the application which can accurately detect, localize and extract text regions in images with complex backgrounds. The resulting system was able to detect, recognize and differentiate similar looking characters and concave curve texts. The proponents also presented the results of the enhancements that they have formulated that made the algorithm efficient in terms of detection of text regions in an image and recognition of similar looking characters and concave curve texts. Text detection algorithm used in character recognition in images is an important approach used to achieve multimedia content retrieval. The proposed algorithm is based on the addition of new steps to the existing algorithm such as the conversion of image to tagged image file format, thresholding and the application of MODI or Microsoft Office Document Imaging. Our proposal is robust enough to accurately localize text regions, effectively recognize concave curve texts and perfectly differentiate similar looking characters.;BACHELOR OF SCIENCE IN COMPUTER STUDIES MAJOR N COMPUTER SCIENCE.;Thesis (Undergraduate) Pamantasan ng Lungsod ng Maynila, 2015. 56

5

5