A Novel Thresholding Method for Text Separation and Document Enhancement

Many thresholding-based image enhancement techniques have been developed and used for document analysis, where the simplicity and efficiency of thresholding makes it ideal to use for classifying layers within documents. However, the efficiency of these enhancement techniques can be impaired by the variation of grey levels in different documents, thus causing over-thresholding or under-thresholding. This paper presents a novel global singlestage thresholding method for separating background and foreground layers in text documents. The method finds an optimum thresholding value or exact separation point for each document using the relationship between luminance value and mean intensity of the document without considering peak values in the grey level histogram. The proposed method is implemented using 50 historical documents and five specifically designed words, and then compared to three other efficient and known thresholding methods. Experimental results suggest that the proposed method performs well for text separation and enhancement of document images.