Answers are grounded in this paper and cite their sources.
Noise-estimation-based fuzzy clustering methods enhance the robustness of fuzzy C-means for noisy image segmentation by modeling residual disturbances. However, existing methods require manual specification of the number of prototypes and random initialization, which can lead to unstable optimization. This paper presents an automatic noise-estimation-based fuzzy C-means framework for image segmentation. An improved density peak algorithm automatically determines the number of prototypes using a Legendre polynomial kernel, while a human mental search strategy ensures robust initialization of the cluster prototypes. Furthermore, the proposed framework uses a superpixel-based residual-driven formulation to reduce computational cost and weighted ℓ2-norm regularization to improve robustness against mixed or unknown noise. On the evaluated real-world images, ANFCM obtains numerically higher mean values than those of SFFCM for segmentation accuracy, Sørensen–Dice similarity, and Matthews correlation coefficient. However, the overall Wilcoxon signed-rank tests reveal no statistically significant differences between the two methods in these three metrics (p=0.6469, 0.5601, and 0.6829, respectively).
No takes yet. Share an insight, caveat, or question.
Wang et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: