On segmentation of images having multi-regions using Gaussian type radial basis kernel in fuzzy sets framework

摘要

Segmentation of images having multi-objects with intensity inhomogeneity and noise is always challenging. In this paper, we propose a new model for segmentation of images having multi-objects with varying intensity. In the proposed model we develop a novel kernel metric which is based on generalized averages. To ensure its applicability in noisy images we use Gaussian type radial basis kernel. To speed up the convergence and to get global optima of the proposed model, we express energy functional of our model in fuzzy Pseudo level set formulation. The proposed model works well in images having multi-objects with intensity inhomogeneity and noise. Our proposed model also works very well in images having maximum, minimum or average intensity background. Instead of length term we use Gaussian smoothing for regularization of Pseudo level set (fuzzy membership function). Experimental results show better performance of the proposed model over existing state of the art models qualitatively and quantitatively (Jaccard similarity). (C) 2017 Elsevier B.V. All rights reserved.