DE19706268A1 - Data-format specific compression process for video and sound data - Google Patents

Abstract

The process is employed in a computer system to reduce the amount of memory used. The compression process is based upon a combination of the method of Huffmann/Lempel-Ziff, run length coding, the difference computation and a pattern recognition. When stored, a data header is used that will be read and a reconstruction takes place.

Description

The invention relates to a data compression and storage method accordingly the preamble of claim 1.

In order to save space and time, one can put computer files with packing programs in one Convert less memory intensive format. Above all, the packing rate important.

Conventional programs (e.g. PKZIP, U.S. Patent 5,051,745) mostly use Coding techniques according to Huffmann / Lempel-Ziff (replacing characters with Bit codes, whereby characters that appear more frequently get a shorter bit code, a longer one) and run length coding (abbreviation longer Sequence of repeating characters). These methods are the ones to be compressed However, file format types are blind to them, so they don't work when they are packaged format-specific (redundant) peculiarities.

The object of the invention is to provide a method that at least some Compresses file types better by taking advantage of their properties makes.

This object is achieved by a method with the features of claim 1.

By means of the file extension, the file headers and an examination of the The process determines the contents of the file as one method of packing the file Combination of Huffmann, run length coding, difference calculation (consideration the differences in the byte values of successive characters) and Pattern recognition.

So the data is not treated equally with a universal procedure, the process is tailored to the different files.

In the case of sound files in particular, the method is far from the conventional methods think.

Due to the principle limitation to just a few packing processes, however can be combined with each other, it is avoided that the Decompression algorithm, which must also be available, becomes too extensive as to what actual packing rate would decrease.

Because in the header only the bit lengths of the Huffmann codes for each character stored (and not traditionally the codes themselves) and from them the Huffmann tree is reconstructed, you also save space.  

The invented process determines the file-specific packing method, writes the Archive header and compresses the data as described below:

A) Packing method

The data is always different from the original state using the difference method converted into five different methods based on the following.

• 1) no change to the original;
• 2. The first byte of the data is left the same, the following bytes are expressed as differences from their predecessors. That has, for example Sound data means that small numbers (distances) occur more frequently and so that it can be compressed particularly well with the Huffmann method.
• 3) Analogous to 2); however, the differences are not direct consecutive bytes are calculated, but rather the differences between one byte and the next. There will be two Difference series formed: one for bytes in odd positions and one for bytes in even positions. These rows are not mixed, but rather written in sequence so that the procedure of run length coding For example, be used as effectively as possible with 16-bit sound data can.
• 4) Analogous to 3), but with three series of differences (each with the differences of bytes in positions with three residues 0, 1 and 2). This method is suitable for example with 24-bit bitmaps (Truecolor).
• 5) Some file types such as TTF (True Type) or WPD (Word Perfect) have sections according to the scheme
  [ASCII 0] [ASCII x] [ASCII 0] [ASCII y] [ASCII 0]. . .
• This is abbreviated to
  [ESC Char. *] [ASCII 0] [ASCII x] [ASCII y]. . . [-ESC Char. *]
  * Escape character: Additional character for which a Huffmann code is also determined.

The area between [ASCII 0] and the second [ESC Char.] Is marked with one of the methods 1) to 4) treated further.

After the conversion, the most suitable of the five in terms of packing rate becomes Process selected using an estimation routine.

The above procedure can be shortened by the file type (e.g. 24-bit BMP, uncompressed) is determined based on file extension and header and is assigned to one of the three packing modes using a predefined table.

During the Huffmann coding, the process of Run length coding applied, where here simply as an escape character further Huffmann code is defined (the number of characters increases if necessary to 257). It is also parallel to Huffmann coding ongoing chunk detection (abbreviation of frequently occurring strings possible with additional Huffmann codes).  

B) Archive header

After calculating the Huffmann tree (according to the known Huffmann algorithm, cf. R. Sedgewick, Algorithms, ISBN 0-201-06673-4) the bit length of the Huffmann code is calculated for each character and from it again constructed a tree using a build tree procedure. This new one Tree is used to compress the file. It is just as efficient like the original tree, since the lengths of the Huffmann codes for each Characters have not changed; however, the code lengths require less Storage space than the individual codes themselves Read lengths from the header and then with the same procedure Build Tree a Huffmann tree constructed so that the archive is unpacked uniquely can be.

The lengths are saved in the archive header as follows:

0100000 00011
Characters | Code length
8 bit * | 5 bit *

* Length can vary, is specified in the header beforehand.

In order to further reduce the memory requirement, runs are abbreviated from characters that follow one another directly in the ASCII character set; instead of

[A] [Code length A] [B] [Code length B] [C] [Code length C] is written shorter:

[C] [A] [Code length A] [Code length B] [Code length C]

The header also contains the file name, an identifier, a checksum and the code number of the packing procedure (necessary for the Unpack) saved.

C) data compression

The original data were used to determine the best packing method converted to one of the five methods previously described; now for each Characters of the converted data into the previously determined Huffmann code Archive written.

D) Unzip

The header of the archive is read in, the packing method is determined from it and reconstructed the Huffmann tree; Now step by step through Undo the compression process the original data restored (each Huffmann code is replaced by the one it describes Characters replaced).

Claims (5)

1. Data compression, characterized in that the method is matched to the file to be packed. 2. The method according to claim 1, characterized in that the packing method as a combination of the compression procedure according to Huffmann / Lempel-Ziff, the Run length coding, difference calculation and pattern recognition is determined. 3. The method according to claim 1 and 2, characterized in that due to file extension, header and - content is searched for redundant information. 4. The method according to the preceding claims, characterized in that image and sound files are exploited the color depth or quality are compressed. 5. The method according to the preceding claims, characterized in that in the file header instead of the individual Huffmann codes only the respective lengths are stored.