Arithmetic Coding

From MaRDI portal





The paper discusses a dual family of so-called ``arithmetic codes for data compression and develops a decodability theory for them. These codes are best viewed as multiple radix representations of sequences as numbers: either greater or smaller than one. These codes collapses to ordinary ``concatenation codes when the radices are picked in a special way. The outstanding feature of arithmetic codes is that they never require alphabet extension, which is an expensive device even in simple applications and impossible to use in complex ones. With arithmetic codes coding for data compression purposes is done in the same general manner as transforming, say, integers in decimal notation to their binary representation; i.e., by adding to the previous code a suitable product of the radices for each symbol to be encoded. Because the numbers involved are generally huge special attention must be paid to how to perform the basic arithmetic operations in normal size registers. The paper discusses how this can be done using certain length parameters as the basic design parameters, these being approximates of the logarithms of the radices.












This page was built for publication: Arithmetic Coding

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4190552)