摘要
主要从传统FDK算法的改进和数据并行计算两方面来研究快速三维图像重建算法,提出了一种Z线优先重建法,能够有效地组织和划分重建数据,从而使得对重建数据的内存访问非常连续,便于采用单指令多数据(Single Instruction Multiple Data, SIMD)技术进行数据并行处理。最后基于Intel Pentium 4 CPU的PC平台,利用SSE/SSE2技术开发了三维图像快速重建引擎。实验结果表明本文提出的方法非常有效,与原始重建算法相比,在保证图像质量不受损失的前提下取得了20倍以上的重建加速比。
We proposed a new kind of fast 3D image reconstruction method that processes Z-Line data first based on improvement on the traditional FDK algorithm and data parallel computing. This method could organize and partition the reconstruction data effectively, which enables accessing memory continuously and makes it easy to implement the SIMD technique on data parallel computing. At last we developed a fast 3D image reconstruction engine on the Intel Pentium 4 platform, using the SSE/SSE2 technique. The result shows that our algorithm is very effective and we can get a reconstruction speedup of over 20 times without any image precision loss compared with the traditional implementation.
出处
《系统仿真学报》
CAS
CSCD
2004年第11期2486-2489,共4页
Journal of System Simulation
基金
国防基础科研项目(J1600E003)
总装预先研究项目(41318.1.1.7)