An Entropy-Based Binarization Method to Separate Foreground from Background in Document Image Processing
Parthajit Roy, Swati Adhikari · SSRN Electronic Journal · 2018
Character recognition is a branch of pattern recognition, where texts or characters are realized from images. Binarization is an important step of character recognition, where foregrounds or texts or characters are separated from backgrounds. The paper proposes a novel entropy-based fuzzy clustering method for text binarization of a gray-level document image. The proposed method considers the spatial information of the gray pixels and the original gray level values for its computations. In order to separate text or document pixels and background pixels in a more realistic way, the proposed method incorporates fuzzy logic-based decision system. The performance of the method has been tested with several images consisting of different Indian languages and the results are evaluated with respect to two existing standard binarization methods.