Binary character/graphic image extraction using fuzzy inference and logical level methods

摘要

Thresholding is one of the most important approaches to image segmentation. It has been widely used to characterize many images by separating objects of reasonably uniform brightness from a background of differing brightness. Typical examples include handwritten/printed texts and microscopic biomedical samples. Despite its broad application to image processing and its theoretical simplicity, there is not, as yet, a robust thresholding technique which can effectively handle the noisy image of nonuniformly distributed brightness. In this study, existing character/graphic image extraction techniques were reviewed and investigated, and new thresholding methods using fuzzy inference and modified logical level are proposed. In particular, in the fuzzy inference technique, new methods applying fuzzification, fuzzy rule and defuzzification are aimed at both lower error and higher processing speed.