A novel algorithm for color quantization by 3D diffusion

Abstract

A novel algorithm for color quantization, three-dimensional frequency diffusion (3D FD), applies to the histogram of an image based on the principle of error diffusion using a 3D error diffusion filter. With the histogram divided into overlapping cubes, an iterative process is devised to select representative colors from these cubes by a popularity scheme that considers a neighborhood of pixels until a colormap is filled. The algorithm is simple but effective. Results show that images quantized by the proposed algorithm are comparable in quality to other popular algorithms.