Color Image Quantization using Interactive Genetic Algorithm

Abstract

Color image quantization is the problem of converting the given image to the image with limited number of colors used, maximizing the similarity. Previous studies measured the similarity as the smaller color difference by pixels, which is hard to evaluate the context of the image. In this thesis, a novel approach of considering the context is shown. Interactive genetic algorithm is applied to reflect the human knowledge in generating the quantized images. The experiment of the system has been done with 12 subjects and four test images, and the result shows that the quantized images generated with the interactive system are evaluated to be more suitable than those with the traditional optimization methods.