On optimized color image coding using correlation of primary colors

Abstract

The RGB color primaries in natural images are characterized by a high degree of inter-correlation. Many compression algorithms use this information redundancy to reduce the amount of bits required for coding, by transforming the color information to a decorrelated color space - such as YUV. The human visual system is more sensitive to luminance than chrominance components, so more bits are allocated to luminance. We examine a different approach, by expressing two of the color components (termed subordinate colors) as a functional of the other color component (termed base color). Unlike some compression algorithms (e.g. JPEG) that perform the analysis on NxN blocks in the image, we utilize segmentation by Region Growing in both gray level and color (RGB) images. We suggest a method for selection of optimal base color for each region separately. The proposed approach could be useful for color compression and progressive transmission applications. © 2013 Springer-Verlag.