Analysis of text-based CAPTCHA images using Template Matching Correlation technique

Abstract

Text-based CAPTCHA images have been widely utilized in on-line applications to anti malicious programs which attempt to make failure in execution or computation. Although installing CAPTCHA enhances system's security, it has to be continuously analysed, improved and developed for hard decoding or extracting from intrusion of automatic programs. This paper is mainly focused on examination of text-based CAPTCHA images with several degrees of noise, skew, font type and size. The Template Matching Correlation (TMC) technique consisting of image conversion, threshold, noise rejection, segmentation and recognition methods, is introduced for analysis. From simulation results, the robustness is increased after the image is distorted by noise background and font skew in the range of 0.3 to 0.4 and 10° to 15°; however fluently recognized by human.