Abstract: The segmentation of the text from the historical degraded image documents is a very challenging task because the variations between the foreground text and background text are hard to determine. In this paper we have implemented a new segmentation algorithm to analyze and extract the words from a degraded image; the process carried here is the image contrast which is adaptively found to solve the issue. Initially the contrast map is taken from the degraded document images. The combination of local image gradient and the local image contrast is the adaptive image contrast, and then it is converted to binary level and combined with canny edge detecting algorithm to extract text edge pixels. The document text is further segmented by a local threshold that is estimated based on the intensities of detected text stroke edge pixels within a local window.

Keywords: Binarization, Otsu threshold, canny edge detector, document enhancement.