Audio Compression Using Fractal Coding

Abstract

Audio Fractal Compression is based on the concept of partitioned iterated function system (PIFS). It exploits the self similarity that is commonly present in audio; this similarity could exploit as a sort of redundancy to compress the audio data. Audio fractal compression finds similar patterns that exist in different scales and different places in audio, and then eliminates as much redundancy as possible.The introduced system consists of two major units; the first is the Encoding unit and the second one is the Decoding unit. In the Encoding unit, the original audio is partitioned into range blocks (non overlapping blocks) and the domain blocks are generated using down sampling with overlapping partitioning, the partitioning step is accomplished using fixed block size audio partitioning scheme. The best matched domain block (i.e., the more self-similar blocks) must be found for each range block by applying an approximate affine transformation. The compression process is finished by storing only the affine transform parameters for every range block. The task of finding self-similarities (via the matching process) is accomplished by making search overall blocks of the domain pool, this will require high computational complexity. This considered a major drawback of the fractal audio compression method. The Decoding unit is typically done by iteratively applying the affine transformations starting with randomly initialized audio data; this transformation repetition is continued until convergence is achieved. The decoding module is less computational demanding than the encoding module. The developed software was tested using four wave samples of data, and it gives encouraging compression result.