Image Segmentation Based on Fuzzy Low-Rank Structural Clustering

Fuzzy clustering is an essential algorithm in image segmentation, and most of them are based on fuzzy c-mean algorithms. However, it sensitive to noise, center point selection, cluster number, distance metric. To address this problem, we propose a new method low-rank representation (LRR) for which integrates structure with theory. First, improve the morphological reconstruction superpixel edge detection by introducing anisotropy enhance edge. Thus, one hand, improved can its noise-resistance performance; other complexity subsequent computation be reduced enhancing superpixels constructed edges. Second, inspired fact that rank represent correlation, concept structure, not dealing data directly but relationship between data. Specifically, perform minimization membership matrix obtain optimal matrix. better results, added Frobenius norm as regularization term LRR model achieve global convergence strong element correlation. Finally, final results processed using subspace constraint. Experiments performed artificial real-world images show proposed more effective efficient than current state-of-the-art methods.