The Extraction of Text/Graphs from Degraded Documents

Abstract

This paper presents a simple and efficient algorithm to clarify the noisy background from a severely degraded document. This method can be applied to enhance the old dated historical documents such that the text to be clear readable condition and the photograph images to be preserved if there is any. We analyze the histogram of the document to find out the characteristics of text-, graph-, and background-pixels. By doing so we identify the graph-pixels and preserve their original gray values and apply Agent Growing Method [1] to the rest of the document in order to clear the noisy background. Our method shows a satisfying result in comparing with several known algorithms.