Fast codes for large alphabet sources and its application to block encoding

For many adaptive codes the speed of coding depends substantially on the alphabet size. In this paper we suggest a method for speeding up codes based on the following main idea. Letters of the alphabet are put in order according to their probabilities (or frequencies of occurrence), and the letters with probabilities close to each others are grouped in subsets (as new super letters), which contain letters with small probabilities. Such a grouping can increase the redundancy of the code. The suggested algorithm is applied to block coding. In order to surmount the block coding problem we suggest applying the described method of grouping to the set of all possible blocks in such a way that the redundancy caused by grouping is relatively small whereas the size of new alphabet is much less than the number of the possible blocks.