RImCom: Raster-order Image Compressor for Embedded Video Applications
| Autor: | Bugra Kertmen, Ozgur Ozkurt, H. Fatih Ugurdag, Okan Palaz, Faruk Donmez |
|---|---|
| Přispěvatelé: | Özyeğin University, Uğurdağ, Hasan Fatih, Palaz, Okan |
| Rok vydání: | 2016 |
| Předmět: |
Computer science
Real-time computing ComputingMethodologies_IMAGEPROCESSINGANDCOMPUTERVISION 02 engineering and technology Lossy compression Theoretical Computer Science 0202 electrical engineering electronic engineering information engineering FPGA Lossless compression business.industry ASIC Frame (networking) Embedded compression Data compression ratio Video processing Bandwidth reduction Memory compression 020202 computer hardware & architecture Video compression picture types Hardware and Architecture Control and Systems Engineering Modeling and Simulation Signal Processing 020201 artificial intelligence & image processing business Computer hardware Information Systems Image compression Data compression |
| Zdroj: | Journal of Signal Processing Systems. 88:149-165 |
| ISSN: | 1939-8115 1939-8018 |
| DOI: | 10.1007/s11265-016-1211-9 |
| Popis: | Due to copyright restrictions, the access to the full text of this article is only available via subscription. This paper presents a real-time, rate controlled, end-to-end (encoder and decoder) hardware solution for memory compression of raster-order video streams—named RImCom (short for Raster-order Image Compression). RImCom offers up to 3x compression that is either lossless or lossy at very reasonable PSNR values. The 180 nm ASIC implementation of RImCom achieves 28 fps at Ultra-HD resolution in the slow corner of synthesis. RImCom can match the fps of the state-of-the-art in the literature with 20 % less area or can achieve twice the fps with 55 % more area. Our FPGA implementation is the only end-to-end FPGA solution in the literature to achieve to this day over 60 fps at Full-HD resolution and to offer rate control. This work was motivated by video processing applications that require the previous frame(s) besides the current frame. When processing HD video streams, even when only one previous frame is required besides the current frame, a significant size and bandwidth of memory is needed. If the current frame is compressed on-the-fly with RImCom or a similar solution and stored on DRAM, and the previous frame is read from DRAM and decompressed with a small IP block, then the overall system cost, power consumption, and electromagnetic radiation are reduced. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|
Full text from SpringerLink