Optimal color image compression using localized color components transforms

Most compression techniques for color images are based on de-correlating the color primaries. Recently, however, a new approach to color image compression, based on exploiting the correlations between the color components, has been presented, outperforming the common de-correlation approach. In this work we introduce an optimized method that generalizes both approaches of correlation and de-correlation and derive its Rate-Distortion (R-D) model. The optimization is based on the selection of localized Color Components Transforms for the compression process. The new method provides results significantly superior to JPEG2000 with respect to all distortion measures and visual quality, while keeping the complexity comparable with JPEG2000.