A novel fuzzy classification entropy approach to image thresholding

In this paper, a novel fuzzy classification entropy approach to generic image thresholding is proposed. Under the assumption that the grayscale histogram of an image follows multimodal distribution, the fuzzy membership function is modified, and the fuzzy entropy is redefined, named fuzzy classification entropy (FCE), to indicate the fitness of the membership function to the actual histogram. The novel membership function and FCE consider not only inter-class distinctness, but also intra-class variety, which provides more accurate description of the histogram. We present bi-level and multi-level thresholding using FCE and conduct experiments on many grayscale images. The results show that the novel method can get moderate thresholds for most images, with better visual quality and less complexity than other fuzzy entropy based methods. (c) 2006 Elsevier B.V. All rights reserved.