Image thresholding method based on Tsallis entropy correlation
Wang, Shaoxun
DOI: https://doi.org/10.1007/s11042-024-19332-3
IF: 2.577
2024-05-12
Multimedia Tools and Applications
Abstract:Image segmentation is an initial task in many vision-based systems and plays an important role in the processes of image analysis, target recognition and tracking, and image thresholding techniques have been widely used because of their simplicity, efficiency and robustness.Tsallis entropy thresholding method is an information-theoretic based thresholding criterion, a global threshold selection method extended from Shannon entropy, which it is assumed that the target and background are independent of each other, and the Tsallis entropy thresholding method may fail if the assumption is not satisfied. To address this, this paper proposes the Tsallis entropy correlation image thresholding method using the Tsallis entropy correlation concept defined by Tsallis, which has the property of generalized entropy and can be applied to the case where the target and the background are not independent of each other. To address the problem that the optimal segmentation threshold computation increases exponentially with the number of thresholds faced by the multi-threshold case, a recursive dynamic programming algorithm under the Tsallis entropy correlation criterion is proposed in the paper. To illustrate the effectiveness of the proposed method, we perform experimental simulations on 20 images from different datasets and compare the image segmentation quality under the original Tsallis entropy, error Tsallis entropy and Tsallis entropy correlation, and the results show that the Tsallis entropy correlation criterion has a good segmentation effect. We also compare the exhaustive method, recursive method, dynamic programming, recursive-based dynamic programming algorithm and particle swarm optimization algorithm to solve the optimal solution of the Tsallis entropy correlation criterion under multi-thresholds, respectively, and the experiments show that our proposed recursive-based dynamic programming algorithm has better stability and lower time complexity, and can effectively solve the optimal solution under multi-thresholds.
computer science, information systems, theory & methods,engineering, electrical & electronic, software engineering