RecView

Outline

RecView is a program for microtomographic reconstruction and image processing. It consists of about 14,000 lines of custom source codes in C++ and CUDA C and x86/x64 machine languages. RecView is designed for processing tomographic data obtained at the BL20B2, BL20XU, and BL47XU beamlines of the synchrotron radiation facility SPring-8.

testPattern Three-dimensional structural analysis by synchrotron radiation microtomography is performed by recording two-dimensional radiographs. Then the three-dimensional structure is determined by the convolution back-projection method. This reconstruction calculation is repeated for each tomographic slice, giving the three-dimensional structure. Therefore, the resolution in the tomographic slice should be affected by the reconstruction calculation. The resolutions in the tomographic slice plane and perpendicular to it have been examined from the modulation transfer functions. The through-plane resolution perpendicular to the tomographic slice corresponded to the spatial resolution of two-dimensional radiographs. By contrast, the in-plane resolution within the slice was approximately 1.5-times as large as the through-plane resolution. Although the two-dimensional radiographs were taken with the pixel width of half the x-ray optics resolution, the three-dimensional resolution analyses indicated that zoom reconstruction should be performed to achieve in-plane resolution comparable to the x-ray optics resolution. Micron 41(1), 90-95. preprint

Licence

RecView is available free of charge to anyone.

References

R. Mizutani, A. Takeuchi, K. Uesugi and Y. Suzuki (2008). Evaluation of the improved three-dimensional resolution of a synchrotron radiation computed tomograph by using a micro-fabricated test pattern. J. Synchrotron Radiation 15(6), 648-654. DOI PubMed pdf
R. Mizutani, A. Takeuchi, R.Y. Osamura, S. Takekoshi, K. Uesugi and Y. Suzuki (2010). Submicrometer tomographic resolution examined using a micro-fabricated test object. Micron 41(1), 90-95. DOI PubMed preprint
R. Mizutani, K. Taguchi, A. Takeuchi, K. Uesugi and Y. Suzuki (2010). Estimation of presampling modulation transfer function in synchrotron radiation microtomography. Nuclear Instrum. Meth. A 621(1-3), 615-619. DOI preprint

Download

The RecView v3 has been released. Download it from here...

The latest RecView release is v3.1.0, currently available from here. This file contains several 'RecView' executables. Please use the 'RecView' executable appropriate for your PC platform. If your PC has NVIDIA Tesla, GeForce or Quadro GPU processors, you can use the CUDA version. The library files 'cudart32_30_14.dll' and 'cufft32_30_14.dll' should be placed in the folder where the RecView CUDA executable was extracted. These library files are also available from the official NVIDIA site as part of the CUDA toolkit. ATI stream processors Radeon, FireStream or FirePro are also supported in this release. To use ATI processors, you should install the OpenCL library 'OpenCL.dll' provided as part of the ATI Catalyst suite.

Release notes

v3.1.0 (released 23 Sep 2011). HIS (Hamamatsu Image Sequence) format is now supported. Source codes for reading HIS format files were kindly provided by Dr. Uesugi, JASRI/SPring-8. Reconstruction routines were revised to support images taken with Zernike phase contrast.
v3.0.1 (released 12 Jan 2011). Least-square fitting functions were provided. Reconstruction routines were revised to support raw image files in the TIFF format.
v3.0.0 (released 27 May 2010). Back projection routines running on Tesla, GeForce, Quadro processors were updated to support the CUDA 3.0 computing environment. The OpenCL 1.0 routines for ATI Radeon, FireStream or FirePro processors were also implemented. Performances are:
- Tau (448 cores, 607 MHz) = 0.0813 nsec (23.4 sec for a 8000x8000 tomogram from 4500 projections) by using GeForce GTX470 with 1.28 GB of GDDR5 memory,
- Tau (112 cores, 500 MHz) = 0.243 nsec (1.75 sec for a 2000x2000 tomogram from 1800 projections) by using Quadro FX 3700 with 512 MB of GDDR3 memory,
- Tau (80 cores, 650 MHz) = 2.86 nsec (20.6 sec for a 2000x2000 tomogram from 1800 projections) by using Radeon HD 5450 with 512 MB of DDR2 memory,
- Tau (4 threads, 3.16 GHz) = 0.825 nsec (5.94 sec for a 2000x2000 tomogram from 1800 projections) by using Xeon (x86),
- Tau (4 threads, 3.16 GHz) = 0.808 nsec (5.82 sec for a 2000x2000 tomogram from 1800 projections) by using Xeon (x64).
Here, tau is the time constant required for one pixel projection. For example, one tomogram of 2000 x 2000 pixels from 1800 projection frames can be reconstructed in tau x 2000 x 2000 x 1800 sec. Drift correction options were also provided in this revision.
v2.0.10 (released 2 Sep 2009). Tilted reconstruction was supported. Angular interpolation was implemented for compatibility with the ct_cbp reconstruction suite. Trimming of marginal regions can be performed in the 8-bit TIFF conversion process.
v2.0.6 (released 13 Mar 2009). Bug-fix release revising memory allocation procedures in the CUDA reconstruction routine.
v2.0.5 (released 9 Feb 2009). Back-projection routines running on Tesla, GeForce or Quadro processors supporting the CUDA programming environment were implemented. The x86 and x64 assembler routines were also revised. Performances are:
- Tau (112 cores, 500 MHz) = 0.314 nsec (2.28 sec for a 2000x2000 tomogram from 1800 projections) using a Quadro FX 3700,
- Tau (3.16 GHz) = 0.633 nsec (4.56 sec for a 2000x2000 tomogram from 1800 projections) using Core2Duo (x86, 2 threads), or tau = 1.336 nsec (x86, single thread),
- Tau (3.16 GHz) = 0.802 nsec (5.77 sec for a 2000x2000 tomogram from 1800 projections) by using Xeon (x86, 4 threads), or tau = 1.724 nsec (x86, single thread),
- Tau (3.16 GHz) = 0.768 nsec (5.53 sec for a 2000x2000 tomogram from 1800 projections) by using Xeon (x64, 4 threads), or tau = 1.651 nsec (x64, single thread).
v2.0.0 (released 21 Jan 2009). x64 platform is now supported. The x64 version can generate larger tomograms upto 10⁶ x 10⁶ pixels. Performance: tau(3.16 GHz) = 0.789 nsec using Xeon (x64, 4 threads). Performace of x86 version: tau(3.16 GHz) = 0.849 nsec using Core2Duo (x86, 2 threads), or tau = 1.395 nsec (x86, single thread).
v1.0.7 (released 23 Aug 2008). Interpolated tomographic reconstruction was implemented. Performance: tau(3.16 GHz) = 1.209 nsec using Core2Duo (x86, 2 threads).
v1.0.4 (released 1 Apr 2008). Multithreaded reconstruction routine was implemented. This function is partially written in x86 machine language, allowing faster execution of the reconstruction calculation.
v1.0.0 (released 6 Mar 2008).

How to use

Concise help can be found in the 'Help'-'About' menu. We believe that this program is self-explanatory, but the following tips should be helpul.

Installation
Download the RecView compressed file. Extract the executable and corresponding library files in any folder you like.

Execution
Double click the 'RecView' executable.

Open radiograph
From the menu bar, select 'File'-'Open...' and choose an ITEX file. Click the 'OK' button.

Computing environment
The GPU or CPU computing environment can be selected in the 'Tomography'-'Property' dialog.

Reconstruction
Open 'Tomography'-'Reconstruction...' dialog. Enter appropriate parameters for your reconstruction calculation. 'Get center' determines the position of the rotation axis. After you have examined a tomogram by using 'Show image', 'Batch' executes the reconstruction calculations.

Trimming or reformatting images
The 'Tomography'-'Histogram=>8bitTIF...' menu provides several tools for generating 8-bit TIFF images.

Frequently asked questions

System requirements.

RecView can be executed on a Windows PC running XP, Vista, or Windows 7 with an x86 or x64 CPU and local storage.

Are any manuals available?

A brief how-to-use guide has been published as the appendix of the following paper:

R. Mizutani, A. Takeuchi, K. Uesugi, S. Takekoshi, R.Y. Osamura and Y. Suzuki (2009). Three-dimensional microstructural analysis of human brain tissue by using synchrotron radiation microtomographs. In Handbook on White Matter, eds. Westland, T.B. & Calton, R.N., New York, Nova Science Publishers, pp. 247-277. pdf (9.5 MB)

What kind of data can be processed?

RecView is designed for the reconstruction of tomographic data obtained at SPring-8. However, any kind of data can be processed. Please contact the author if you wish to use this program with data from other tomographs.

Contact / bug report

Ryuta Mizutani, Dr.
Department of Applied Biochemistry
School of Engineering, Tokai University
Kitakaname 4-1-1, Hiratsuka, Kanagawa 259-1292, Japan
E-mail ryuta(at)tokai-u.jp
HP http://www.el.u-tokai.ac.jp/ryuta/