This paper proposes a new method for reducing the number of colors in an im
age. The proposed approach uses both the image color components and local i
mage characteristics to feed a Kohonen self-organized feature map (SOFM) ne
ural network. After training, the neurons of the output competition layer d
efine the proper color classes. The final image has the dominant image colo
rs and its texture approaches the image local characteristics used. To spee
d up the entire algorithm and reduce memory requirements, a fractal scannin
g subsampling technique can be used. The method is applicable to all types
of color images and can be easily extended to accommodate any type of spati
al characteristics. Several experimental and comparative results are presen
ted. (C) 1999 John Wiley & Sons, Inc.