When compressing color images in pre-press applications one has to deal with four different color channels. Usually there exists a high correlation between the different colors in natural images. In the type of images that will be dealt with in this paper, this correlation is even higher because of the addition of a fourth color component, black, due to various technical reasons. This paper shows that a significant gain in data reduction can be achieved by taking this color redundancy into account. Before performing a lossy block-based compression technique like JPEG, a singular value decomposition (SVD) is performed on the data. Therefore the image is divided into blocks, that usually have a different size than those used in the JPEG step, and the SVD is applied locally to decorrelate the colors in the spatial domain.