Full text (gzip-compressed postscript)
We present a new method for error modeling applicable to the MLP algorithm for hierarchical lossless image compression. This method, based on a concept called the variability index, provides accurate models for pixel prediction errors without requiring explicit transmission of the models. We also use the variability index to show that prediction errors do not always follow the Laplace distribution, as is commonly assumed; replacing the Laplace distribution with a more general symmetric exponential distribution further improves compression. We describe a new compression measurement called compression gain, and we give experimental results showing that the MLP method using the variability index technique for error modeling gives significantly more compression gain than other methods in the literature.