Image segmentation method based on fuzzy entropy and grey relational analysis

Abstract

A new image segmentation method based on grey relational analysis and fuzzy entropy is presented.The traditional fuzzy entropy methods are sensitive to the noise,because they only consider the statistical information of the gray values.In the proposed method grey relatinal degree is introduced to demonstrate whether the pixels belongs to object or background more accurately.The current pixel and its neighbors are selected as a comparative sequence,and the grey relational degree between the comparative sequence and the reference sequence is computed,based on which the membership function of the fuzzy entropy function is redefined so that the membership of the current pixel is determined by its own gray value and the gray values of its neighbor pixels.The segmentation experimental results of several real images exhibit the good performance on reducing the noise by the proposed method.