作者: Jeffrey A. Fessler
DOI:
关键词: Mathematics 、 Computer vision 、 Radon transform 、 Fast Fourier transform 、 Iterative reconstruction 、 Artificial intelligence 、 Fourier transform 、 Computation 、 Iterative method 、 Interpolation 、 Projection (set theory)
摘要: Fourier-based reprojection methods have the potential to reduce computation time in iterative tomographic image reconstruction. Interpolation errors are a limita- tion of methods. We apply min-max interpolation method for nonuniform fast Fourier transform (NUFFT) minimize errors. Numerical results show that NUFFT approach provides substantially lower approximation er- rors and backprojection than Iterative reconstruction offer numerous advantages over conventional filtered method. The late 1990's saw commercial release 2D PET SPECT systems. burden forward operations remains primary hin- drance wider use fully 3D im- age This paper describes new efficient using combina- Fourier-slice theorem transform. is particularly well suited geometries scan- ners. For most methods, each iter- ation requires one "forward projection" "backprojection," where projection roughly discretized evaluation Radon transform, backprojector adjoint projec- tor. steps traditionally