Using inverse image frequency for perception-based color image quantization

The process of selecting a small number of representative colors from an image of higher color resolution is called color image quantization. A well-known problem in quantizing images is to select the best representative colors that not only reduce the quantization error, but also account for the perception of human vision. The technique we propose effectively handles this problem by using the variation of colors in different regions of an image, in addition to the use of the color histogram, for effective perception and quantization. We introduce the property of inverse image frequency (IIF) for computing the representative colors of an image. IIF is based on the observation that colors within a color subset having non-uniform frequency distribution across the different regions of an image have better discriminating properties than those having uniform distribution. Our approach to incorporate the information derived from IIF can be combined with any standard quantization algorithm. The results show that our approach quantizes an image more effectively than using just the well-known median cut algorithm.