A Fast Wavelet Algorithm for the Reduction of Color Information

Color depth reduction is a common operation in the processing of image data, particularly for image file format conversion or for mapping images to display hardware. A problem of particular interest is the mapping of 24-bit deep RGB data to an 8-bit indexed color model i.e. how to choose 256 (or less) representative colors from a set of up to 2 (16.8 million) colors. Solutions to this problem range from the generation of an optimal palette, which is a time-consuming process, to the fast but approximate method of mapping to a regular grid i.e. a color cube. In this paper we develop a practical wavelet-based thresholding algorithm to compute a near optimal palette in close to real time.