Fuzzy ART for Relatively Fast Unsupervised Image Color Quantization

Abstract

The use of Fuzzy Adaptive Resonance Theory (FA) is explored for the unsupervised color quantization of a color image. The red, green and blue color component values of a given color image are passed as input instances into FA which then groups similar colors into the same class. The average of all of the colors in a given class then replaces the pixel values whose original colors belonged to that class. The FA unsupervised clustering is capable of realizing color quantization with competitive accuracy and arguably low computation time.