Image thresholding using LBP and GA-based optimal parameter selection for 2D Tsallis-Havrda-Charvat entropy

In this paper, we propose an automatic thresholding method based on 2D Tsallis-Havrda-Charvat entropy and histogram of local binary patterns (LBP). Tsallis-Havrda-Charvat entropy is extracted from 2D histogram, which is calculated by using the LBP decimal value of a pixel and the average decimal value of its local neighborhood. Few parameters influenced the thresholding results. Therefore, an automatic optimal parameter selection using GA (Genetic algorithm) has been proposed here. Based on the optimal parameter selection for 2D Tsallis-Havrda-Charvat entropy using GA and maximizing the criterion function, we obtain the best possible threshold pair. LBP histogram is adopted to capture the texture information. LBP's high performance for texture characterization helps to make our method more suitable for thresholding the images enriched with texture information. Finally, GA improved the thresholding result by selecting optimal parameters. In this paper, we test the efficiency of our thresholding method when applied to some real-world images, and experiments show that our proposed method is promising and robust in terms of efficiency.