Abstract: The relative entropy thresholding segmentation algorithm based on the co-occurrence matrix is a commonly used image segmentation method. An asymmetric co-occurrence matrix was constructed by an adaptive filter method to improve the relative entropy thresholding segmentation method, which is better adapted to the noise images segmentation. hhe segmentation experiment results show that this method can reduce the noise interferes more effectively, and get the more complete objectives and the more distinct edge.

Key words: Image segmentation, Relative entropy, Co-occurrence matrix, Minimum variance filtering