An Acceleration Strategy for Generating Cone-Beam CT Images Based on "Multi-core" Systems | ||||
Journal of Advanced Engineering Trends | ||||
Public Titles, Volume 40, Issue 2, April 2021, Page 117-126 PDF (650.23 K) | ||||
DOI: 10.21608/jaet.2020.39654.1029 | ||||
View on SCiNiTO | ||||
Authors | ||||
marwa mohamed abbas 1; Hassan Ali Youness 2; Ammar Mostafa Hassan3 | ||||
1Minia University – Faculty of Engineering, Computers and Systems Engineering Dept., Minia, Egypt | ||||
2Minia University – Faculty of Engineering, Computers and Systems Engineering Dept., Minia, Egypt | ||||
3Arab Academy for Science, Technology and Maritime Transport, South Valley Branch, Aswan, Egypt | ||||
Abstract | ||||
This paper displays a speeding up system for a cone-beam generation. We proposed a strategy to accelerate the Feld Kamp, Davis, and Kress (FDK) calculation based on a tiling process which is a compiler enhancement procedure. The test was the way by which to improve the use of the cache by strategies for loop blocking. Wiping out the cache misses as conceivable as possible is the guideline; reason/motivation behind the loop tiling. The method is represented and indicated by which method can be used as part of the recreation code, and in case of recognizing the reproduction output on the CPU "Multicore" condition; a specific model is also used. Exploratory results show that the intended Strategy achieves estimated 8.98 throughput per second (PPS) for reconstructing 〖512〗^3-voxel volume from 360 〖512〗^2 Pixel projections. This execution undertakes 3.37 speedup and 8.9 throughputs over than naive CPU based strategy that represents the infrastructure for parallelization code. Also, error measures results show that the proposed method did not affect the image quality reconstructed by the naïve method. | ||||
Keywords | ||||
FDK; Loop tiling; Cone-beam; Filtered back projection | ||||
Supplementary Files
|
||||
Statistics Article View: 236 PDF Download: 335 |
||||