Spatial-chromatic clustering for color image compression

A relaxation spatial-chromatic clustering for color image compression is presented. It exploits both the spatial correlation and the chromatic contents to the benefit of the fuzzy clustering algorithm by defining the membership function with spatial-chromatic hybrid vector. A relaxation process with splitting and merging is employed to refine the compression results. The constrained random initialization and re-initialization of the codebook vectors help guarantee that the algorithm is not trapped into the local minimum. Experimental results demonstrated that the proposed scheme could efficiently reduce the storage requirement while maintaining high image quality and was able to diminish noise without blurring edges