Topic
Abstract
Infrared images are by nature fuzzy and noisy, thus the segmentation of human targets from them is a challenging task. In this work, fuzzy Tsallis entropy of background and target are defined respectively according to probability division principle. Next, a newly defined entropy is extended into two dimensions to make full use of spatial information. To overcome the huge calculation burden induced by extending one dimensional method into two dimensional one, a fast algorithm of two-dimension fuzzy Tsallis entropy is put forward to reduce the computation complexity from O(L2) to O(L). Finally, the parameters of fuzzy member function are optimized by shuffled frog-leaping algorithm following the maximum entropy principle, and thus the optimal threshold of image is obtained. Compared with typical entropy-based thresholding methods by experiments, the presented method is verified to be efficient and robust.