A Color Image Quantization Algorithm Based on Kernel Method
基于核方法的彩色图像量化研究

CIQ(Color image quantization) refers to the process of replacing N colors in an image with K colors,and making the new image is as much as close to the original one.This paper introduced a KCIQ(Kernel Color Image Quantization) algorithm based on K-means clustering.By using kernel method,data from low-dimensional input space can be mapped to(high-dimensional) feature space,so that no-linear problem may become linear one in feature space.Since this method involves only inner products of data in input space,it can be evaluated much more efficiently even with no knowledge of mapped data in feature space.Our experiments show that KCIQ performs better than K-means on special image,which includes one important color but with only a few data,and KCIQ has almost the same performance as K-means when working on common images.