Neutrosophic C-means clustering with local information and noise distance-based kernel metric image segmentation

Abstract

The traditional FCM algorithm is developed on the basis of classical fuzzy theory, though the classical fuzzy theory has its own limitations. The lack of expressive ability of uncertain information makes it hard for FCM algorithm to handle clustered boundary pixels and outliers. This paper proposes a Neutrosophic C-means Clustering with local information and noise distance-based kernel metric for image segmentation (NKWNLICM). At first, noisy distance and fuzzy spatial information are introduced to NCM model to improve the robustness of noise image segmentation. Then, the kernel function is used to measure the distance between pixels. By mapping low-dimensional data into high-dimensional data, the classification performance is further improved. At last, the fuzzy factor is redefined based on the distance between the center pixel and its neighborhood. The new fuzzy factor can excellently reflect the influence of neighborhood pixels on central pixels and improve the classification accuracy much better. The experimental results on Berkeley Segmentation Database demonstrates the excellent performance of the proposed method for noisy image segmentation.