With our compression method we have optimised arithmetic coding, which gives the best compression ratio, so it is up to 30% quicker, which saves you time compressing data.
At the moment, the state of the art consists of creating large compressed archives, which have to be decompressed as a whole before they can be worked with. With our method this is not necessary anymore, because despite the fact that the data is compressed, this method is able to access single records and even smallest information units in short time accurately. This saves you a lot of time, where you normally would wait for the compression/decompression of your data; because only the really needed data is decompressed.
Each data compression method can be divided – as latest scientific discoveries show - into at least three components. The Densifier™ data compression software implements latest technologies in each component which are carefully optimized to work with each other in synergy. In contrast, common compression systems implement ad-hoc conglomerates of these three components which lead to archives that are larger than necessary and which are neither reusable in different combinations nor open to general improvements from the research community:
1. Modeling: The splitting of the input data in data packages - so-called symbols, which are compressed separately. In this way a text can for example be divided into words, syllables, or single characters. A picture can be divided into many little squares with 3x3, 4x4, 5x5, and so on pixels, into textures, color gradients, and other graphical elements.
2. Statistics management: Counts how often a symbol has appeared up to a certain point in time or evaluates the probability with which this symbol is likely to occur. These two figures can usually be converted. This kind of statistics management is dependent on the coding component.
3. Coding: Determines the probability of symbols occurring in bit streams. Two practical methods exist:
a) Huffman coding: The quickest but less compact coding method.
b) Arithmetic coding: The theoretically and practically most compact coding method (for this module; the modeling component can be optimized separately), because arithmetic coding attains the data entropy with its compression density exactly. Unfortunately, arithmetic coding is much slower than Huffman coding.
Practice has shown that the increased computing time for arithmetic coding pays off if symbols with high probabilities are common. This is particularly true for the compression of structured data.
Modeling is the application-specific component for which no optimal compression schemes exist. Unfortunately it is generally impossible to optimize in favor of compression density, compression speed, and main memory usage at the same time. These three characteristics can in the general case only be traded off against each other. What makes the Densifier™ data compression unique is that much effort went into making these optimizations for many specialized domains by using knowledge about the structure of the input data or by learning this knowledge automatically in an intelligent way. These “knowledge-based” specializations and resulting optimizations can only be applied to the modeling component, which is therefore the heart of the Densifier™ data compression innovation.
Also our accelerated method of statistics management for arithmetic coding, as well as accelerations for arithmetic coding and Huffman coding can be practicably applied to any kind of compression method.
Each individual data compression application can be optimized in terms of speed or compression density — control elements that we can also set to optimally suit the needs of your current project.
If you have a specific data compression task, we should together identify the most frequent data types (graphics, texts, structured data, etc.) with their characteristics. Of course we’ll gladly help you with this and with the selection of well-suited compression methods. If you have many different data types, you should either use all specialized compression methods or mainly the general method.
The data throughput in the data sheet is generally specified in kBytes per second. 1 kB/second equals 8 kBits per second (kbps) or 0,0078 Mbps (Megabits per second). These transfer ratios relate – unless otherwise specified – to the amount of data prior to compression.
The removed percentage of the original file size that is removed by compression: 1 – compressed size / uncompressed size. A compression ratio of 1:10 is equivalent to a removed percentage of 90%. The larger this value the better the method is for compressing. Other terms like "compression rate" are very ambiguous and thus are avoided.
Information &Questions: firstname.lastname@example.org
|www.compris.com |||Data Compression |||Contact/Map |||About Compris Intelligence GmbH|