Dynamic model selection during data compression

• US 5,045,852 A
• Filed: 03/30/1990
• Issued: 09/03/1991
• Est. Priority Date: 03/30/1990
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Mitchell, Joan L., Rissanen, Jorma J., Pennebaker, William B.
• Primary Examiner(s): Pellinen, A. D.
• Assistant Examiner(s): Hoff, Marc S.
• Application Number:
  US07/502,909
• Time in Patent Office: 522 Days
• Field of Search: 341/51, 341/65, 341/67, 341/95
• US Class Current: 341/51
• CPC Class Codes: H03M 7/4006 Conversion to or from arithmetic code
• CPC Class Codes: H03M 7/48 alternating with other codes during the code conversion process, e.g. run-length coding being performed only as long as sufficientlylong runs of digits of the same kind are present

Abstract

A system and method for maximizing data compression by optimizing model selection during coding of an input stream of data symbols, wherein at least two models are run and compared, and the model with the best coding performance for a given-size segment or block of compressed data is selected such that only its block is used in an output data stream. The best performance is determined by 1) respectively producing comparable-size blocks of compressed data from the input stream with the use of the two, or more, models and 2) selecting the model which compresses the most input data. In the preferred embodiment, respective strings of data are produced with each model from the symbol data and are coded with an adaptive arithmetic coder into the compressed data. Each block of compressed data is started by coding the decision to use the model currently being run and all models start with the arithmetic coder parameters established at the end of the preceding block. Only the compressed code stream of the best model is used in the output and that code stream has in it the overhead for selection of that model. Since the decision as to which model to run is made in the compressed data domain, i.e., the best model is chosen on the basis of which model coded the most input symbols for a given-size compressed block, rather than after coding a given number of input symbols, the model selection decision overhead scales with the compressed data. Successively selected compressed blocks are combined as an output code stream to produce an optimum output of compressed data, from input symbols, for storage or transmission.

Claims

1. A computerized method of maximizing the compression of a stream of input symbols by dynamically selecting the best of a plurality of data compression models for respectively encoding each of successive portions of said stream, wherein each model, in response to input symbols, generates encoded data, based on model characteristics, in the form of respective strings of data, comprising the steps of:

beginning with the same input symbol in said stream, using each model to generate respective strings of data from the stream of input symbols;

compression coding said strings of data to produce blocks of compressed data, each of said blocks being at least equal to a given size;

selecting the respective block of compressed data for which the most input symbols have been compressed;

adding the selected block of compressed data to an output data stream; and

beginning with the input symbol following the last symbol input to said selected model, reiterating the foregoing steps on the next succeeding portion of said input symbol stream.

2. The method of claim 1 wherein the compression coding step comprises arithmetic coding.

3. The method of claim 1 wherein the compression coding step comprises binary arithmetic coding and said blocks of compressed data are only completed after a less probable symbol has been coded.

4. The method of claim 1 further comprising the steps of:

coding a model identifier at the start of each block of compressed data indicative of the model producing the respective block; and

including the model identifier for the selected block therein when adding the selected block of compressed data to the output data stream.

5. The method of claim 4 wherein said model identifier is used to decode the compressed data block.

6. The method of claim 1 wherein the past history of said respective strings of data is used for identifying the model to be used for the next block of compressed data.

7. The method of claim 6 wherein a threshold value is selected based on the number of input symbols coded per compressed data block, and said model to be used for the next block of compressed data continues to be used, and alternate models are not processed, as long as said threshold value is exceeded by the number of input symbols coded per compressed data block.

8. The method of claim 1 wherein said blocks of compressed data have probability estimation values associated therewith and further comprising the step of resetting the probability estimation values for the models not selected to the previous estimation values at the start of the production of the next block.

9. The method of claim 1 wherein one of said models comprises comparing the size of the selected block of compressed data to the size of the number of input symbols that have been compressed and selecting the smaller of the two for adding to the output data stream.

10. The method of claim 1 wherein the compression coding step comprises variable length coding.

11. A computerized method of selecting one of a plurality of data compression models for compressing data symbols in a stream of data symbols to maximize compression of the stream, comprising the steps of:

using each of said plurality of models to form a respective block of compressed data from data symbols input from the stream, said forming beginning with the same data symbol in said stream and ending with said respective blocks of compressed data being of comparable size; and

selecting the model which produces the block, from among said respective blocks, for which the most input symbols have been compressed.

12. The method of claim 11 wherein said blocks of compressed data are formed by the steps of:

beginning with the same input symbol in said stream, using each model to generate respective strings of data from the stream of input symbols; and

compression coding said strings of data to produce said blocks of compressed data.

13. The method of claim 12 wherein said respective blocks of compressed data are only completed after a less probable symbol has been coded.

14. A computerized method of dynamically selecting the best of a plurality of data compression models for use in arithmetic coding and decoding wherein each model generates a string of data in response to a stream of input symbols, comprising the steps of:

beginning with the same symbol in said input symbol stream, using each of said models to produce respective strings of data from said input symbols including a model selection decision for the respective model;

coding the model selection decision and the string of data for each model with an adaptive arithmetic coder to generate respective blocks of compressed data for each model, with each of said blocks being of at least a given size and starting the coding of each model for a given block with the same code stream and probability interval; and

selecting the model with the best coding performance from among the blocks of compressed data generated during the arithmetic coding on the basis of the amount of input symbol data compressed.

15. A computerized system for maximizing the compression of a stream of input symbols by dynamically selecting the best of a plurality of data compression models for respectively encoding each of successive portions of said stream, wherein each model, in response to input symbols, generates encoded data, based on model characteristics, in the form of respective strings of data, comprising:

means, beginning with the same input symbol in said stream, for generating respective strings of data from the stream of input symbols using each of said data compression models;

means for compression coding said strings of data to produce respective blocks of compressed data, each of said blocks being at least equal to a given size;

means for selecting the respective block of compressed data for which the most input symbols have been compressed;

means for adding the selected block of compressed data to an output data stream; and

means for actuating said foregoing means to act on the next succeeding portion of said input symbol stream, beginning with the input symbol following the last symbol input to said selected model, until all of the portions of said input stream have been compressed and added to said output data stream.

16. The system of claim 15 wherein said compression coding means comprises means for arithmetically coding said strings of data.

17. The system of claim 15 wherein said compression coding means comprises means for variable length coding said strings of data.

18. The system of claim 15 further comprising means for resetting the probability estimation values, associated with said blocks of compressed data, for the models not selected to the previous estimation values at the start of the production of the next block.

19. A computerized system for selecting one of a plurality of data compression models for compressing data symbols in a stream of data symbols to maximize compression of the stream, comprising:

means, using each of said plurality of models, for forming a respective block of compressed data from data symbols input from the stream, said forming beginning with the same data symbol in said stream and ending with said respective blocks of compressed data being of comparable size; and

means for selecting the block, from among said respective blocks, for which the most input symbols have been compressed.

20. A computerized system for dynamically selecting one of a plurality of data compression models for use in arithmetic coding and decoding wherein each model generates a string of data in response to a stream of input symbols, comprising the steps of:

means, beginning with the same symbol in said input symbol stream, for producing respective strings of data from said input symbols using each of said models and including a model selection decision for the respective model;

adaptive arithmetic coder means for coding the model selection decision and the string of data for each model to generate respective blocks of compressed data for each model, with each of said blocks being of at least a given size and starting the coding of each model for a given block with the same code stream and probability interval; and

means for selecting the model with the optimum coding performance from among the blocks of compressed data generated during the arithmetic coding on the basis of the amount of input symbol data compressed.