Algorithms and hardware implementation of high-performance data/image compression
| Vortragende(r): |
| Alfredo de la Cruz Nogueiras (GEMAC Chemnitz) |
| Inhalt: |
| Data compression algorithms are representative examples of iterative hardware algorithms, but still most part of existing solutions have been restricted to software environments, imposing several limitations concerning performance. In fact, the goal of reaching truly on-line compression decompression is actually quite demanded in many distributed communication and storage solutions, traducing itself in advantages in storage capacity and improved bandwidth. Still, requirements imposed by hardware design implies a re-thinking of suitable resources, memory consumption and overall strategy in the compression algorithm. In this presentation, we discuss the design of the RFGD novel data compression solution, expressed and tested as IP-Cores for other ASIC designs. The compression algorithm developed shows, as main advantage against competing solutions, substantial improvements in compression rates, as confirmed by tests from leading companies in the storage industry. A comparative analysis is done against equivalent solutions to underline the reduced resources employed, at the same time reaching performances of more than 33 MBytes/s (or 256 Mbits/s) to solve most on-line applications. Additionally, it is discussed the overall strategy employed to extend the use of RFGD to multimedia data types. In specific, it is shown how, by usign predictive Image filters as a preprocessing unit, lossless and quasi-lossless compression rates of images are highly improved. An RFGD-based solution for hardware image compression is shown, allowing to work on-line and reaching compression rates quite superior to the ubicuous lossless TIF format. |
| Zeiten: |
| Mittwoch, der 12.01.2005, 11:00 Uhr, Raum 1/305 |
