Beating the Quality of JPEG 2000 with Anisotropic Diffusion

Although widely used standards such as JPEG and JPEG 2000 exist in the literature, lossy image compression is still a subject of ongoing research. Galić et al. (2008) have shown that compression based on edge-enhancing anisotropic diffusion can outperform JPEG for medium to high compression ratios when the interpolation points are chosen as vertices of an adaptive triangulation. In this paper we demonstrate that it is even possible to beat the quality of the much more advanced JPEG 2000 standard when one uses subdivisions on rectangles and a number of additional optimisations. They include improved entropy coding, brightness rescaling, diffusivity optimisation, and interpolation swapping. Experiments on classical test images are presented that illustrate the potential of our approach.