|Supervisor:||M.Sc. Andreas Heindel (Room 01.178)|
|Faculty:||Prof. Dr.-Ing. André Kaup|
Digital image coding algorithms are present in many areas of today's life. A very high compression ration is achieved for so-called consumer applications, where lossy compression is employed. However, many professional applications require lossless compression which comes with significantly higher data rates.
An compromise between lossy and lossless compression is lossless scalable coding. Here, reconstruction quality of a lossy compression base layer may be improved using a losslessly compressed enhancement layer. Depending on the application it is then possible to decode only one layer or both layers.
For efficient compression of the enhancement layer, the so-called SELC scheme has been proposed. As basic version of SELC is already available as a dedicated codec, allowing various lossy codecs like JPEG, JPEG 2000, or H.264/AVC intra coding being used for the base layer.
In this Master's thesis, the current implementation shall be extended by not yet implemented features of SELC. First, a more efficient processing order of the enhancement layer pixels still has to be considered. Second, the entropy coding should be extended by so-called context models. The processing order (e.g. the respective base layer block raster) as well as the use of the context models shall be adapted to the different used base layer codecs. Finally, the coding efficiency as well as the runtime of the scalable coding system is to be evaluated.
The implementation has to be done in C++. For that reason at least some experience in C is a required. Basic knowledge of image and video compression, especially entropy coding, is advantageuous.