Iterative multi-level image segmentation based on fuzzy set theory

To solve low accuracy and time-consuming problems in thresholding, an iterative multi-level image segmentation method based on fuzzy set theory was proposed. The interactive top-down region dividing strategy was used considering characteristics of gray level distribution and neighborhood information of each pixel in the pre-dividing region. And fuzzy comprehensive evaluation was used instead of segmentation thresholds to classify each pixel in the image reasonably. The method overcomes the limitations and complexity in thresholding, and improves the quality and efficiency of image segmentation. To test the superiority and rapidity of presented method, a set of noisy images and different contrast images, along with two representative complex images, were segmented. And two classical thresholding methods, Otsu's method and Kapur's method, were used as comparisons. The experimental results show that the presented method costs shorter time and preserves detail information better for the concerned.