Paths to future image lossless coding

In the paper introduction of non-MMSE (minimum mean square error) based predictor optimization techniques to adaptive lossless image compression is considered. The proposed cost function is Minkowski distance, being generalization of MMSE and MMAE (minimum mean absolute error). It is proven empirically and explained why MMSE criterion is suboptimal. Then, currently the best lossless image coding methods are outlined (AVE-WLS, Blend-20), and introduction of new criteria to their definitions discussed.