Supplemental Materials: Deep denoising for multi-dimensional synchrotron X-ray tomography without high-quality reference data
Supplemental Materials: Deep denoising for multi-dimensional synchrotron X-ray tomography without high-quality reference data
复制标题
补充材料:在没有高质量参考数据的情况下对多维同步加速器 X 射线断层扫描进行深度去噪
DOI:
--
复制
发表时间:
2021
期刊:
影响因子:
--
通讯作者:
K. Batenburg
中科院分区:
文献类型:
--
作者:
A. Hendriksen;Minna Bührer;Laura Leone;M. Merlini;N. Viganò;D. Pelt;F. Marone;M. Michiel;K. Batenburg
X-ray tomography datasets were acquired at the TOMCAT beamline at the Swiss Light Source (SLS) at the Paul Scherrer Institut (PSI), Villigen, Switzerland. In addition, an X-ray diffraction tomography (XRD-CT) dataset was acquired at the ID15A beamline at the European Synchrotron (ESRF), Grenoble, France. Static X-ray 3D micro-tomography The data was acquired at the TOMCAT beamline1, and is publicly available2. The dataset contained 1001 projection images across an angular range of 180° measuring 1100×1440 square pixels of size 2.75 μm. The acquisition took 1s, the exposure time was 1ms, and the mean energy of the polychromatic beam was 30keV. Dynamic X-ray micro-tomography This dataset was acquired at the TOMCAT beamline1, and is publicly available2. A full scan was performed every 0.1s, resulting in a dataset of 180 time steps spanning approximately 32 seconds, divided into three 6s chunks with 7s pauses in between. Each time step contained 300 projection images across an angular range of 180° measuring 1100×1440 square pixels of size 2.75 μm. The mean energy of the polychromatic beam was 30keV. The documented angular increment was π/300 radians, consistent with the acquisition of 300 projection images per half rotation2. We observed a small deviation in this case. The angular increment was π/299.924, which could be determined up to five significant digits by visual inspection of the combined GridRec reconstruction of several time steps. XRD-CT High-energy X-ray diffraction measurements were taken at the ID15A beamline using a monochromatic pencil beam (90keV energy). Data was collected of 3 horizontal slices, spaced 7mm apart, and acquisition of each slice took 20 minutes. Acquisition was performed in 273 translation steps over a scan range of 12mm and in 225 rotational steps over an angular range of 180°. Sinograms were computed from the acquired images using the pyFAI library3. A subset of the sinograms was selected, containing 3 horizontal slices with 11 channels each. The displayed FBP reconstructions were computed with the Shepp-Logan filter using the ASTRA-toolbox4.