Rank order polynomial decomposition for image compression

In this paper, a novel decomposition scheme for image compression is presented. It is capable to apply any nonlinear model to compress images in a lossless way. Here, a very efficient polynomial model that considers spatial information as well as order statistic information is introduced. This new rank order polynomial decomposition (ROPD) that allows also for a progressive bitstream is applied to various images of different nature and compared to the morphological subband decomposition (MSD) and to the best prediction mode for lossless compression of the international standard JPEG. For all compressed images, ROPD provides better compression results than MSD and clearly outperforms the lossless mode of JPEG.