Arithmetic Compression

Densifier®: Arithmetic Coding

Arithmetic Coding is the theoretically best-packing compression method. An aspect which can still be improved is the compression speed. This has been done by Compris Intelligence GmbH.

Decoding up to 50% quicker
Because of our new module for statistics management for arithmetic coding, the decoding is up to 50% quicker.
The coding is about as fast as known from previous approaches.
Therefore, for example, a cheap 333 MHz AMD K6-processor can have as good a decompression performance as an Intel 400 MHz MMX Pentium II processor, which is twice as expensive, using the old method.

Our module can easily be integrated in already existing data compression systems. Because most of the data is compressed once, but decompressed many times (e.g. data in the internet / on servers or in other archives), it is especially important to have a module which can decompress very fast. Also, it is possible to decompress more quickly data which has been compressed with other systems using the arithmetic format.

Fast and accurate access to single data records and even smallest information units is possible in spite of compression. Only the information really needed is decompressed.

Densifier™ data compression module for increasing the speed of arithmetic encoding & decoding (we do also speed-improved versions of Huffman coding)



typical Values

Speed up of the statistics-module
for encoding


Speed up of the statistics module
for decoding


Speed up of the total
arithmetic compression process


Speed up of the total
arithmetic decompression process


Memory requirements

no additional memory needed; no additional memory for administrative information besides 1-4 Bytes per frequency-value

Operating systems

MS-Windows NT, 2k, XP, 2003, Unix (Linux, Solaris); others as requested


can optimize most existing compression programs by replacing less efficient modules

Further procedure

a)    Customer uses the quasi standardized  call interface with object code provided by

b)    Customer supplies function heads necessary for calling the statistics gathering or coding module. will supply the customized component for that interface as object code.

Additional characteristics / advantages:

·        Presumably the fastest method for arithmetic coding/decoding worldwide

·        even data compressed with other variants of arithmetic coding can still be decompressed more quickly

·        a little slower than Huffman coding, but attains the theoretical & practical best compression density

·        support and update service as individually agreed upon

Available programming components


·        Software development kit with documentation in English

·        US $2900,- once per platform for all components, variable costs for the library per installed computer: US $50,-
– no patent-license costs


All data compression systems can be broken down into three components. The modeling module contains the user specific behavior: These are audio, video and graphics compression or general text and binary compression. Depending on the kind of modeling component, many different possible compression densities can be obtained. For the combination of statistics & coding, a theoretical & practical most optimal solution exists: Arithmetic coding. In this module only probabilities are transformed into bit streams, ""e"" for example occurs with a probability of 5% in English. Huffman coding is a ""quick and dirty"" variant: Huffman coding is distinctly faster, but its compression density is also distinctly worse.


Compris also offers a variant of the Huffman coding with increased operation speed. Conditions regarding prices/licensing are equivalent to the ones given above for arithmetic coding.

Information & Questions:  | Data Compression  | Contact/Map  | About Compris Intelligence GmbH