Volume 16, No 2, 2019
Image Segmentation Using Elbow Embedded Rough Fuzzy K-Means
Amit Kumar Mandal
Abstract
In this paper, an Elbow embedded Rough Fuzzy K-means (ERFKM) approach for image segmentation is proposed. K-means requires advanced knowledge of K value and is dependent on initialization of centroids, which is random initialization trap. In this approach, these limitations are overcome by using Elbow method to select number of cluster (K) and by rough set’s reduction concept to initialization of centroids in Fuzzy K-means (FKM). First we use Elbow method to select the value of K, choose the cluster centroids and then apply rough set theory (RST) to reduce and optimize these centroids. After that we apply FKM on the reduced and optimized centroids set to segment the image. FKM is used modeled the fuzzy membership functions to determine similarity of image’s pixels. Experimental results show that ERFKM outperforms FKM both visually and theoretically in terms of segmentation, clustering time and Xie-Beni Index.
Pages: 381-391
Keywords: Elbow, Rough Set, Fuzzy Set, K-means, Xie-Beni Index.