Reclamation of information from degraded and damaged document images by image binarization method
B. Vishnudharan, K. Anusudha · 2017
Reclamation of information from severely degraded document images is a highly difficult job because of huge dissimilarity among the document contextual surface (background) and the forefront text, characters and shapes of the various document images. This paper proposes an eminent document image binarization method which focusses these problems with the aid of an adaptive map. This adaptive map is a combo of a contrast image and the gradient image, which is resist to the dissimilarity present among the document background and the forefront text/characters of various document images suffering from various degradation and damages such as irregular lighting, image contrast dissimilarity and ink bleeding-through. In this proposed method, initially an adaptive map is created for each and every input degraded document images. Then binarization takes place to that contrast map to determine the boundary edges of the text stroke pixels by fusing it with canny edge map. The forefront text present in the document is divided by means of local thresholding, which predicts the forefront text depending upon the detected text stroke edge pixels intensity value. Considerable tests were conducted on the degraded document images and the performance were evaluated based on parametric such as PSNR and MSE respectively. The performance analysis indicates that this system shows promising results against a wide variety of document degradations.