Meat Image Segmentation Using Fuzzy Local Information C-Means Clustering for Generalized or Mixed Kernel Function
Article
Figures
Metrics
Preview PDF
Reference
Related
Cited by
Materials
Abstract:
Focusing on the lack of a strong adaptive ability to noise of the meat image segmentation method that is based on the traditional kernel fuzzy C-means (KFCM) clustering, two image segmentation methods (KFLICM_UG and KFLICM_MG) were applied to meat samples; these segmentation methods used fuzzy local information C-means clustering (FLICM), which is based on generalized or hybrid kernel function. Firstly, the generalized or hybrid kernel functions were used to strike a good balance between learning and generalization abilities. Each image pixel was mapped onto a high-dimensional feature space, which leads to a larger inter-class difference between the useful features of pixels. Thus, those pixels could be clustered more easily in the high-dimensional feature space. Then, the fuzzy membership of each pixel was determined based on the combination of its local space information with grayscale information. Finally, meat image segmentation was completed via the fuzzy local information C-means clustering according to the image features in the high-dimensional feature space. Considering results of previous studies, compared with the existing FCM (Fuzzy C-Means) segmentation methods such as KFCM and FLICM segmentation methods, the proposed method (KFLICM_UG, KFLICM_MG) can achieve better segmentation results with lower segmentation error rate, stronger adaptiveness, and robustness against noise.
