An efficient framework for lossless color image compression

There are many methods for compressing color images nowadays and most of them are lossy. However, in many important situations, lossless color image compression is irreplaceable. In this paper, we propose an efficient lossless color image compression framework. In this framework, for an RGB image, it is first decorrelated via a reversible color transform. In the new color space, the color components are directly down sampled and then interpolated to the original size. Next we subtract these interpolated components from their original counterparts to obtain the prediction errors, which together with the color subcomponents are compressed via Huffman coding. At the decoder, the exact reverse process is used to reconstruct the original color images. Experimental results show that the proposed method achieves overall better compression performance compared with the famous CALIC algorithm and some other popular methods used in the TIFF and PNG color image compression standards.