Adaptive Color Quantization Based on Self-Growing Network

Color quantization is an important technology in visual information processing. A new algorithm for color quantization is proposed which automatically estimates number of the representative colors to efficiently represent an arbitrary image. It is based on the growing mechanism of Growing When Required neural network and a novel method to visit uniformly pixels in an image. A number of criteria are introduced that have an effect on controlling of the number and topology of neurons in the output layer. The experiments demonstrate that the developed method can automatically define the number of representative colors while keeping distortion to an acceptable level. It is also shown that the algorithm outperforms the popular ones in terms of color distortion.