Title :
Data partitioning for parallel implementation of real-time video processing systems
Author :
O´Nils, Mattias ; Lilljefjäll, Per-Robert ; Thörnberg, Benny
Author_Institution :
Dept. of Inf. Technol. & Media, Mid Sweden Univ., Sundsvall, Sweden
fDate :
28 Aug.-2 Sept. 2005
Abstract :
The main reason for using data partitioning for a video processing system is the increase in performance it offers. This paper presents a comparison of the buffering requirements for different data partitioning methods used in video processing systems. The analysis shows that the buffer storage required for the implementation of video systems is highly dependent on the partitioning strategy. The results indicate that partitioning the tasks vertically is by far the most efficient method when considering buffer sizes.
Keywords :
buffer storage; data compression; memory architecture; parallel processing; real-time systems; video signal processing; buffer size; buffer storage; data partitioning; memory access; memory structure; parallel implementation; power consumption; real-time video processing system; Buffer storage; Convolution; Cost function; Design optimization; Energy consumption; Image analysis; Image processing; Information technology; Performance analysis; Real time systems;
Conference_Titel :
Circuit Theory and Design, 2005. Proceedings of the 2005 European Conference on
Print_ISBN :
0-7803-9066-0
DOI :
10.1109/ECCTD.2005.1522948
