Deep neighborhood structure driven interval type-2 kernel fuzzy c-means clustering with local versus non-local information

Chengmao Wu,Siyun Peng

DOI: https://doi.org/10.1007/s11042-023-15230-2

IF: 2.577

2023-04-15

Multimedia Tools and Applications

Abstract:For images with high noise, existing robust fuzzy clustering-related methods are difficult to obtain satisfactory segmentation results. Hence, this paper proposes a novel single fuzzifier interval type-2 kernel-based fuzzy local and non-local information c-means clustering driven by a deep neighborhood structure for strong noise image segmentation. Based on the neighborhood window around the current pixel, we firstly construct the novel deep neighborhood window structure, which is composed of neighborhood window around the current pixel and neighborhood window around pixels in the neighborhood window around the current pixel. Secondly maximally and minimally neighborhood weighted distances between current pixels and clustering centers are obtained through deep neighborhood window structure. Thirdly, two local neighborhood distances are used to modify upper and lower fuzzy membership of robust single fuzzifier interval type-2 fuzzy clustering with kernel metric and local versus non-local information, and an enhanced robust interval type-2 kernel-based fuzzy clustering with single fuzzifier is presented for strong noise image segmentation. Experimental results indicate that the proposed algorithm has better segmentation performance and stronger anti-noise robustness, and outperforms existing state-of-the-art robust fuzzy clustering-related algorithms in the presence of high noise. In particular, the segmentation accuracy of the proposed algorithm is 20% higher than that of the important KWFLICM algorithm and has increased by 10% compared with the FALRCM algorithm proposed in recent years.

computer science, information systems, theory & methods,engineering, electrical & electronic, software engineering