원문정보
초록
영어
In this paper, a new three-level thresholding method for image segmentation is proposed based on nonextensive entropy and fuzzy sets theory. Firstly, the image histogram is transformed from crisp set to fuzzy domain using fuzzy membership function, such as triangular membership function. After that, the nonextensive entropy of each part of fuzzy domain of histogram is computed. The threshold is selected by maximizing the nonextensive fuzzy entropy. However, the search of combination of membership function’s parameters is costly. For reduce the computation time, the artificial bee colony algorithm is used to search the optimal combination of the membership function’s parameters. The experimental results on tested images demonstrate the success of the proposed approach compared with the competing methods.
목차
1. Introduction
2. Thresholding Principle Based On Fuzzy Set and Nonextensive Entropy
2.1. Image Transformed From Crisp Set to Fuzzy Set
2.2. Thresholding through Nonextensive Fuzzy Entropy
3. Threshold Selection Using Artificial Bee Colony Algorithm
4. Experiment Results and Analysis
4.1. Performance Evaluation
4.2. Experiments on Real Images
5. Conclusion
Acknowledgements
References