Multi-Threaded Filtered Backprojection Algorithm on Multi-Core Processors
Yun Ho Oh, Won Woo Ro · ICEIC : International Conference on Electronics, Informations and Communications · 2010
In this paper, we introduce the parallelized filtered backprojection algorithm for multi-core processors. To exploit parallelism and achieve better performance, two parallelized implementation methods are presented and their execution times are compared. Of these approaches, the method which parallelizes the matrix operation has no data dependency so that extra operation is not needed. With this advantage, it shows better performance than the strategy which is based on the loop-level parallelization. The maximum speedup is achieved as 6.73 compared to the sequential execution. These results are 78% faster than the parallelization method which divides the workload by the outermost iteration of the backprojection-interpolation stage.