Robust Image Segmentation Using KFCM with Noise Restrained

An image segmentation algorithm based on pulse-coupled neural network is presented in this paper. Our main motives of using the kernel methods consist in: kernel functions can be viewed as a non-linear transformation that increases the separability of the input data by mapping them from the feature space to a new high dimensional space. In this paper, at first, we introduce kernel function which enables the c-means algorithm to explore the inherent data pattern in the new space. Secondly, a new metric is proposed in KFCM to replace the conventional Euclidean norm by introducing the concept of weight, we claim that the proposed new metric is more robust than the conventional c-means clustering algorithm for the case when the data are damaged by some noise. The kernel-based fuzzy k-means clustering algorithm with noise restrained enhances robustness of the original clustering algorithms to noise. Experimental results with artificial and real-world images have shown the effectiveness of the proposed algorithm