Abstract

In this article, we propose a modified Livewire algorithm, Generalized Fuzzy Livewire (GFLW) that is different from classical Livewire algorithm, to segment color and texture image, such as medical image. The key of Livewire is how to build the weight matrix. We characterize image using splitting and merging scheme on J-value criteria proposed by Deng firstly, and transform the color or texture image to scalar value image. As the J-values at pixels neighboring real boundary usually are close, which mean fuzzy, it is unsuitable to build weight matrix on J-value image. Our scheme is to express the J-value image with a suitable generalized fuzzy set. And then we process this generalized fuzzy set by generalized fuzzy set theory to build the weight matrix. Based on this weight matrix, we segment image using the shortest path algorithm. The experiment results show the modified algorithm is useful.

Paper Details