Reversible integer approximation of color space transforms for lossless compression of big color raster data

Abstract

Reversible integer transforms are or great importance for lossless compression algorithms. To perform reversible decorrelation of color channels we propose an algorithm for calculating parameters of a reversible integer transform, which approximates such continuous mappings as a discrete Karunen-Loeve transform. We propose a method for estimating the approximation errors, which makes it possible to choose an optimal approximation of the original transform that minimizes these errors. Using the MRG file format, intended for storing large amounts of integer raster data, as an example, we show that after applying decorrelation, it is possible to increase the compression ratio of multichannel raster images using lossless compression algorithms.