Robust dynamic semi-supervised picture fuzzy local information clustering with kernel metric and spatial information for noisy image segmentation

Aiming at robust picture fuzzy clustering with weak anti-noise ability, which is difficult to meet the needs of high noise image segmentation. Hence, this paper proposes a robust dynamic semi-supervised picture fuzzy local information clustering with kernel metric and spatial information. Firstly, a robust weighted squared Euclidean distance between the current pixel and clustering center is constructed, and it is introduced into picture fuzzy clustering to form a robust spatial picture fuzzy clustering. Secondly, the membership degree of neighborhood pixels is linearly weighted to form local membership of the current pixel, which is used to supervise the picture fuzzy partition information of spatial picture fuzzy clustering, and a novel robust dynamic semi-supervised picture fuzzy clustering with spatial information is obtained. Subsequently, the fuzzy local information factor is embedded in dynamic semi-supervised spatial picture fuzzy clustering to obtain enhanced dynamic semi-supervised picture fuzzy local information clustering. Finally, to further improve the adaptability of dynamic semi-supervised spatial picture fuzzy local information clustering, a robust dynamic semi-supervised picture fuzzy local information clustering with kernel metric and spatial information constraints is proposed. Experimental results show that the proposed algorithm outperforms existing state-of-the-art robust picture fuzzy clustering-related algorithms for high noise image segmentation, and the Acc and PSNR indexes of the proposed algorithm in noisy images are significantly better than those of robust fuzzy clustering-related compared algorithms.