Transformation coding

from Wikipedia, the free encyclopedia

Transform coding (English transform coding is) a method of data compression for signals such as audio signals (1D) or images (2D). The actual data compression is usually preceded by a frequency transformation . This has the advantage that the resulting coefficients have a statistically significant distribution and can therefore be better modeled and subsequently compressed.

In other words, after the transformation, some coefficients are predictably larger, others smaller. Therefore, one can neglect smaller coefficients ( lossy compression ) and / or use entropy coding ( lossless compression ).

The majority of the common modern lossy compressing audio and video formats such as the MPEG format family ( MP3 , MPEG-2 , ...) are based on transform coding using the modified discrete cosine transform (MDCT), which is a form of the fast Fourier transform (FFT). represents. However, especially in the area of ​​image processing, attempts are being made to work with wavelet transformations.