New Approach to Image Segmentation Based on Neighborhood-Influenced Fuzzy C-Means Clustering

In recent years, accurate segmentation of images is a very challenging task for image processing applications. Image segmentation can also be stated as clustering problem in which image pixels are clustered according to the homogeneity of their feature values. Crisp K-means clustering algorithm can achieve the solution of this problem. But it is not suitable for coinciding partition and it is unable to handle noisy data. Fuzzy form of C-means clustering can manage overlapping partition problem, but traditional FCM is also sensitive to noise pixels. In this paper, neighborhood-influenced Fuzzy C-means (NFCM) algorithm is proposed where spatial neighborhood information of pixels is incorporated with the traditional fuzzy c-means algorithm. NFCM is giving more accurate segmentation result compared to hard c-means and fuzzy c-means based segmentation techniques.