An efficient adaptive Masi entropy multilevel thresholding algorithm based on dynamic programming

Masi entropy multilevel thresholding can utilize additive and non-extensive information in images to effectively segment a complex image. However, the entropy index of Masi entropy cannot be selected automatically, and the time complexity of the multilevel algorithm by exhaustive searching grows exponentially with the increase of the threshold numbers. To address these two problems, an adaptive entropy index selection strategy based on image histogram information is proposed first. To improve the computation efficiency, an efficient solution for the adaptive Masi entropy multilevel thresholding algorithm based on dynamic programming (DP + AMasi) is also proposed. The DP + AMasi algorithm is compared with the Masi entropy multilevel thresholding algorithm by exhaustive search and state-of-the-art metaheuristic algorithms on three benchmark datasets. The effectiveness of the DP + AMasi is verified by fitness function values, Uniformity Measure, Davies Bouldin index, and CPU run time. In addition, the Wilcoxon test is used to analyze the differences between algorithms.