A Novel Fuzzy C-Means Clustering with Hybrid Local and Non Local Spatial Information for Brain Magnetic Resonance Image Segmentation

Abstract

Due to the poor performance of image segmentation algorithms with the local spatial information in an image heavily contaminated by noise, some non-local spatial information considered for fuzzy c-means algorithms in recent years. But it seems more reasonable taking into account both local and non-local spatial information because usually there are valuable information about each pixel in its adjacent neighbourhood. In this paper a novel fuzzy c-means clustering algorithm with hybrid local and non-local spatial information (FCM_HLNL) is proposed. The FCM_HLNL utilizes both local and non-local spatial information in order to make segmentation robust encountering high level noise in images. In the proposed method, a novel nonlocal adaptive spatial constraint term is used to modify the objective function of fuzzy c-means algorithm. The characteristic of this technique is that the adaptive spatial parameter for each pixel is designed to make the hybrid local and non-local spatial information of each pixel playing a positive role in guiding the noisy image segmentation. Segmentation experiments on synthetic and real images, especially brain magnetic resonance (MR) images, are performed to assess the performance of an RFCM_HLNL in comparison with other and fuzzy c-means clustering algorithms with local and non-local spatial constraint. Experimental results show that the proposed method is robust to high level noise in the image and more effective than the other comparative algorithms.