Performance Analysis of Different Kernel Functions for MRI Image Segmentation

Abstract

Nowadays, Kernel-based clustering methods have gained popularity over the widely used Euclidean distance-based methods in the area of image segmentation. Kernel-based methods generally replace the inner product with the positive definite function and perform a non-linear mapping of input data into a high dimensional feature space. The use of these non-linear kernel methods makes the process of clustering more robust and accurate. Different types of kernel methods have been used for the process of clustering. In this paper, we are going to explore the performance of different kernel functions applied with fuzzy c-means for the process of image segmentation. We have incorporated Gaussian kernel function, hyper-tangent kernel function, log-based kernel function and Cauchy kernel function. Experiments are performed on various synthetic and real MRI brain images. The performance is measured on the basis of segmentation accuracy, false-negative ratio and false-positive ratio.