New statistical models of the JPEG lossless mode subject to the super high definition images

This paper describes new statistical models of the JPEG lossless mode subject to the super high definition images (SHDI). Seven predictors prepared in the JPEG are very simple to alleviate the complexity of the prediction process, which indicates that prediction residuals correlate. The actual correlation of the residuals exhibits a tendency to be more significant as the number of picture elements increases. Consequently, the conditional probability densities of the residual signals for SHDI differ from the Laplacian distribution commonly assumed in predictive coding. We propose two statistical models considering the peculiar probability densities and investigate the validity of the models by coding simulations