Lossless image coding based on minimum mean absolute error predictors
Yoshihiko Hashidume, Yoshitaka Morikawa · 2007
For prediction-based lossless image coding, the coding performance depends largely on the efficiency of predictors. In general, mmse predictors are well used, but these predictors suffer from large errors at edges. In response, the authors have proposed minimum mean absolute error (mmae) predictors which are less sensitive to edges. Mmae predictors provide accurate prediction and entropy of prediction errors is reduced. In this paper we infer prediction errors based on mmae and mmse predictors can be modeled by the Laplacian and Gaussian function, respectively, and conclude mmae predictors are superior to mmse predictors in terms of coding performance.