Document Images Binarization Using Hybrid Combination of Fuzzy C-Means and Deghost Method

Abstract

This paper presents a document binarization approach to the document image. Thousands of historical documents usually hold important information within them. It is usually stored in the national archives and library around the globe waiting to be scanned to retrieve the content it holds. However, many environmental factors, improper handling, and the poor quality of the materials used in the document creation cause it to suffer a high degree of degradation of various types. In order to retrieve the content of the degraded historical document, a binarization approach on document images must be applied. The processes of the binarization separate pixel value of an input image into two values which is white as the background and black as foreground text. The proposed system consists of three parts. In the first part, the image pre-processing operation is done before the binarization process to enhance image quality. In this part, Contrast Stretching and Mean Filter is applied onto the image to remove noise on the image. The second part will be to apply the binarization algorithm on the document image that has undergone an image pre-processing operation. By applying the Fuzzy C-Means algorithm to the document images, the images will be converted to a binary image and divided into two components, which is text and background. The last step of the proposed method will be performing the Deghost operation to remove "ghost" entities that may have appeared on the document image. The method will undergo imaging quality analysis, such as PSNR, Accuracy, and F-measure to determine the effectiveness of the proposed method. The experimental results on H-DIBCO 2013 dataset show the robustness, reliability, and efficiency in the proposed approach.