Discrimination of handwritten and machine printed text in scanned document images based on Rough Set theory

Discrimination of handwritten and machine printed text in a scanned document image is an important phase as processing and recognizing machine printed and handwritten text cannot be done using a single OCR. In this paper a novel approach has been proposed to discriminate machine printed and handwritten text using Rough Set theory. The uniform occurrence of characters in a word is considered as the main feature for discrimination. Uniformity has been depicted from the transitions occurred due to the overlay of component structures on the null background. Rough sets are used to build the knowledge of uniformity among the characters in the word. Based on the equivalence relation between the defined rough set and the derived set, words are identified as machine printed or handwritten. Extensive experiments have been conducted on locally generated 400 samples and samples from IAM dataset.