A Novel Approach Based on Hybridization of Fuzzy C-MEANS and Competitive Agglomeration for Image Segmentation

The proposed work was aimed to evaluate the hybridization of fuzzy C- means and competitive agglomeration for image segmentation. It is fuzzy clustering-based vector quantization algorithm. The method of segmenting a digital image into multiple segments is known as image segmentation. For segmentation, various methodologies based on standard techniques such as region rising, threshold technique, and watershed transform have been proposed. Because of the drawbacks of these approaches, segmentation based on clustering techniques has emerged. The central idea behind data clustering is that each cluster is represented by its centroid. In addition, the similarity of the input vectors to the centroid is used to represent each cluster. Clustering approaches are divided into two categories: parametric and nonparametric. Finding natural groupings in a dataset using a Euclidean distance between samples is a non-parametric technique. K-means, hierarchical, and spectral clustering are examples of non-parametric clustering. The lack of robustness to image noise is one of these methods' drawbacks. As a result, in image clustering and segmentation, a fuzzy segmentation technique has been commonly used. The most challenging aspect of fuzzy c-means is determining the number of clusters and choosing an objective function. As a result, we use a fuzzy clustering-based vector quantization algorithm. This algorithm makes use of a specialised objective function that combines fuzzy c-means and a competitive agglomeration term. This algorithm is fast, and the reconstructed images are of good quality.