Novel fuzzy clustering algorithm with variable multi-pixel fitting spatial information for image segmentation. (January 2022)

• Record Type: Journal Article
• Authors: Zhang, Hang; Li, Haili; Chen, Ning; Chen, Shengfeng; Liu, Jian
• Is Part Of: Pattern recognition. Volume 121(2022)
• Publication Date: 2022-01
• Subjects: Fuzzy clustering -- Image segmentation -- Spatial information -- Variable filter window -- Variable generalized neighbourhood window; Pattern perception -- Periodicals; Perception des structures -- Périodiques; Patroonherkenning; 006.4
• DOI: 10.1016/j.patcog.2021.108201
• Languages: English
• ISSNs: 0031-3203

Abstract

Abstract: Spatial information is often used to enhance the robustness of traditional fuzzy c-means (FCM) clustering algorithms. Although some recently emerged improvements are remarkable, the computational complexity of these algorithms is high, which may lead to lack of practicability. To address this problem, an efficient variant named the fuzzy clustering algorithm with variable multi-pixel fitting spatial information (FCM-VMF) is presented. First, a fuzzy clustering algorithm with multi-pixel fitting spatial information (FCM-MF) is developed. Specifically, by dividing the input image into several filter windows, the spatial information of all pixels in each filter window can be obtained simultaneously by fitting the pixels in its corresponding neighbourhood window, which enormously reduces the computational complexity. However, the FCM-MF may result in the loss of edge information. Therefore, the FCM-VMF integrates a variable window strategy with FCM-MF. In this strategy, to preserve more edge information, the sizes of the filter window and generalized neighbourhood window are adaptively reduced. The experimental results show that FCM-VMF is as effective as some recent algorithms. Notably, the FCM-VMF has extremely high efficiency, which means it has a better prospect of application.