/rectv_gpu

Four-dimensional tomographic reconstruction by time domain decomposition. Version on GPU

Primary LanguageC++BSD 2-Clause "Simplified" LicenseBSD-2-Clause

rectv_gpu

Four-dimensional tomographic reconstruction by time domain decomposition

Installation

Building python modules

Set CUDAHOME environmental variable, run

python setup.py install

Simple reconstruction scenario

Read, filter, normalize data and save it to file 'data.npy'

python read_continuous

Reconstruct with the time-domain decompositon + regularization python rec_simple.py

Use as a module

See an example in tomobank https://tomobank.readthedocs.io/en/latest/source/data/docs.data.dynamic.html#foam-data

python tomopy_rectv.py dk_MCFG_1_p_s1_.h5 --type subset --nsino 0.75 --binning 2 --tv True --frame 95

--type - reconstruction type (slice,subset,full)

--nsino - location of the sinogram used by slice or subset reconstruction (0 top, 1 bottom)

--binning - factor for data downsampling (0,1,2)

--tv - use tv reconstruction (True,False)

--frame - central time frame for reconstruction, 8 time frames will be reconstructed by default. Example --frame 95 gives time frames [91,99)