On selection of centroids of fuzzy clusters for color classification
Dae-Won Kim, Kwang H. Lee
TL;DR
The paper tackles the initialization sensitivity of fuzzy c-means for color clustering by introducing a dominant-color initialization strategy. It builds a color-domain membership model to 14 ColorChecker reference colors in the perceptually uniform CIELAB space, ranks reference colors by their maximum observed membership, and selects the top c dominant colors to seed the initial centroids as the closest data points to these colors. The approach integrates a fuzzy reference-color membership with standard FCM updates, and demonstrates, via a 10-point example, how dominant colors (e.g., Black, Red, Yellow) guide centroid placement. This method aims to improve convergence reliability and clustering quality in color segmentation tasks by leveraging perceptual color distinctions.
Abstract
A novel initialization method in the fuzzy c-means (FCM) algorithm is proposed for the color clustering problem. Given a set of color points, the proposed initialization extracts dominant colors that are the most vivid and distinguishable colors. Color points closest to the dominant colors are selected as initial centroids in the FCM. To obtain the dominant colors and their closest color points, we introduce reference colors and define a fuzzy membership model between a color point and a reference color.