Spectral clustering based text image segmentation using fuzzy logic

A novel approach of text images segmentation based on fuzzy logic and spectral clustering is proposed in this paper in order to extract the text from complex backgrounds. Parameters of the function are obtained according to the maximum entropy principle and the original image is fuzzified. Then gray, distance and textural information among pixels are extracted from the fuzzified image to construct the affinity matrix. The original image is segmented using the clustered eigenvector corresponding to the minimum eigenvalue of the matrix. Experimental results show that the proposed method is superior to the usual thresholding methods and can handle the natural scene text image with complex backgrounds.