Preprocessing techniques for improving the lossless compression of images with quasi-sparse and locally sparse histograms

Among the characteristics found relatively frequently in computer-generated images, but that are usually not found in natural images, is intensity histogram sparseness. The difficulties shown by state-of-the-art image coding algorithms in properly compressing images with sparse histograms have been pointed out in some recent works. In this paper, we address not only the problem of compressing images belonging to this class, but also the problem of compressing images that, although not possessing histograms that are strictly sparse, can be classified as quasi-sparse or locally sparse. We propose some simple preprocessing techniques that may lead to some dramatic improvements in the compression ratios attained by state-of-the-art image coding techniques.