Genetic Algorithm Clustering for Color Image Quantization

Clustering is an unsupervised classification method used for different issues in image analysis. Genetic algorithms are randomized search and optimization techniques. In this paper, we present a genetic algorithm clustering for color image quantization as a prior process to any other one for image analysis. A fitness function with a smallest number of variables is proposed. It’s based on the fuzzy c-means objective function reformulated by Bezdek and the one proposed by Frigui and Krishnapuram in their competitive agglomeration algorithm. In the proposed clustering genetic algorithm, variable chromosome length is used to adjust the clusters number and determine their centers at the same time. Application of the algorithm to color image quantization shows that initial population solutions converge to good results. Furthermore, results in different color spaces and impact of parameters are discussed.