HST_MASTER Program Options

After the file sequence of projection images has been defined TOMOGRAPHY RECONSTRUCTION CONTROL FORM 1 appears. A typical example of this form is shown in Figure 5, Page .

  

Figure 5: Form 1

By clicking on the yellow right-hand column of buttons the ``values'' of the variables controlling the reconstruction can be changed. When the ``values'' have all been set appropriately the O.K button can be clicked to move to the second control form.

The basic sense of each control parameter is described in the left-hand column of DESCRIPTIONS. Further explanation is given when the value is being changed.

The parameters are described following the button text:

FILE INTERVAL   This control parameter allows faster, but poorer quality reconstruction by only using one in so many images for the reconstruction. This option would not normally be used, except for ``quick-look'' reconstructions to check the reconstruction volume. For full reconstruction of all images, this item should equal 1. To only use every other image this item should equal 2. To use every third image this item should equal 3, etc.

This item could also be used for sequences where the number increment between images is higher than one.

-BACKGROUND    

This control may be YES or NO. If the value is YES a background image will be selected and subtracted from each data image prior to flat-field correction and other operations, and subsequent back projection. This is normally performed as detectors, and in particular CCD detectors, have a non X-ray dark current signal which must be subtracted to put the data on a linear scale.

(The background image is also subtracted from any flat-field image prior to it being used for normalisation.)

FLAT-FIELD    

This parameter determines whether or not data files will be flat-field corrected prior to further correction and back-projection. It may take the value YES or NO. If YES one or more files will be selected for the flat-field images.

Again, whilst efforts are made to ensure that the detectors have as uniform a response to X-ray intensity as possible, and that the X-ray beam is as uniform as possible, there are always imperfections, and reconstructed data quality is much improved by correcting for these imperfections.

TAKE LOG    

This parameter controls whether or not the log of the (corrected) data values is taken prior to back-projection, and may have the value YES or NO. Normally the value would be set to YES so that the absorption  of the sample is calculated. However, if the data has already been treated outside of the program then it may be required not take the logarithm and this value should be set to NO.

VERT. ROTATION      

This parameter defines whether the axis of rotation was vertical (YES) or was horizontal (NO). This is defined relative to the detector read-out.

For correct reconstruction and for the calculation of the position of the rotation axis it is vital that this is set correctly.

(It would be possible to have a horizontal rotation axis but to rotate the detector by 90 degrees   and still have this value set YES. This will result in slightly faster reconstruction as it is more efficient in reading the detector files from disk.)

CACHE KILOBYTES    

This parameter defines the size of the processor cache (L2) per processor in kilobytes. This value is used to to decide whether or not to process more than one slice at the same time. If the cache is large enough it may be more efficient to process two slices at the same time. However, if the cache is not large enough this might result in slower reconstruction owing to the need to write data to and from the RAM.

For the ESRF computer systems the following values are correct:

``alpha'' (Compaq XP-1000 workstation): 4096

``beta'' (Compaq DS-20 server): 8192

Typical PC system: 512

SINOGRAM MEGABYTES  

This is the maximum size of storage in megabytes that will be used for storing internally the sinograms prior to back-projection. Normally this value should be set lower than the physical memory size (RAM). If this value is set too high the system will not have enough RAM to store the sinograms internally, virtual memory will have to be used, and the reconstruction will take longer. This can be monitored by looking at disk usage . Normally the disk usage should be very low and periodic, corresponding to the output of reconstructed slices. If the disk is being used continually than the system is not running well, reconstruction will take longer, and eventually the disk will wear out and breakdown. Under these circumstances you should reduce this value for future reconstructions. If this value is set lower than necessary it means that the reconstruction will be broken down into a larger number of passes during which a smaller number of sinograms will be stored. This is less damaging, but nevertheless means that the reconstruction will take longer than is necessary.

For a system with 1024 Mbyte of RAM a value of about 800Mbytes should be appropriate. Ideally, RAM should be left for the system, other arrays within hst_slave, and perhaps for other processes.

OUTPUT SINOGRAMS    

This parameter allows the assembled sinograms to the output as a series of files, one sinogram per file. For normal full reconstruction this would be set to NO. However, it is sometimes necessary to pre-treat    the sinogram data prior to full reconstruction in which case this item may be set to YES.

RECONSTRUCTION    

For normal reconstruction and output of the reconstructed volume this value should be set to YES. If just the sinograms are to be assembled and output then this value can be set to NO and no back-projection will take place.

After clicking on O.K. on form 1 TOMOGRAPHY RECONSTRUCTION CONTROL FORM 2 appears. A typical example of this form is shown in Figure 6, Page .

  

Figure 6: Form 2

This works in a similar fashion to form 1, but the control parameters are different.

The parameters are described following the button text:

OVER-SAMPLING   hst_slave can reconstruct efficiently whilst maintaining reconstruction data quality, by over-sampling the detector pixels, and by linearly interpolating   within the over-sampled array. Having done this, the appropriate over-sampled detector pixel which is back-projected into the reconstructed volume can be found without any further need for interpolation. This allows the reconstruction to be almost as fast as a simple nearest pixel algorithm, whilst allowing the data quality of the interpolation method to be maintained. (In fact interpolation methods other than linear could also be applied in this manner with minimal increase in overall reconstruction time.)

This parameter determines the accuracy with which interpolation is applied to back-project into voxel positions which fall between two detector pixel positions. The value 0 means that full linear interpolation will be used; this will result in the reconstruction taking considerably longer time. The value 1 means that very simple nearest pixel  values will be used and no interpolation will take place. This will result in the quickest reconstruction but quality may suffer. Values greater than one means that the detector pixel arrays will be over-sampled and linearly interpolated at the over-sampling factor specified. The reconstruction then takes place using the closest over-sampled pixel. This allows reconstruction almost as fast as nearest pixel reconstruction but with much better data quality, depending on of course using a suitable over-sampling factor. For most data the value of 4 seems to be quite adequate, however even if a larger value needs to be used this will still be faster than linear interpolation for every pixel.

INCREMENT ANGLE   

This is the incremental angle in degrees from one data image to the next. It may be expressed in degrees and fraction of degrees. (The program works in single precision arithmetic, so you may usefully express this value to 6 or 7 significant figures.)

The numbering of the first and last image in the projection file series is used to suggest the default value for this parameter. This assumes the last image is just before the 180 degree rotation image.

OUTPUT FILE  

This parameter is the name of the of file to contain the completed reconstruction. By default it is the same as the first projection file, but with the extension changed to vol.

If you want to change this file name, click of the button, and the file selection utility will allow you to change output directory and select the required file name.

X-PIXEL SIZE  

This is the pixel size in the faster changing direction (horizontal) in microns. The pixel sizes are used to to convert the absorption  into physical units, but are otherwise unimportant. This parameter will be used for data-sets where the rotation axis was vertical (relative to displayed images).

Y-PIXEL SIZE  

This is the size of the pixels in the slower changing direction (vertical) in microns. This parameter will be used for data-sets where the rotation axis was horizontal (relative to displayed images).

OFFSET ANGLE    

This parameter allows the object reconstructed to be arbitrarily rotated by an angle relative to the experimental conditions. This value is expressed in degrees and may also contain fractions of degrees.

OUTPUT GRAPHICS    

If the value is set to YES then during reconstruction, periodically reconstructed slices will be displayed. This may be used to monitor data quality and check that the reconstruction is progressing in a reasonable manner.

Now hst_master calculates the image closest to 90 degree rotation . This is displayed and you can asked to select a rectangle to reconstruct, in the same manner as for the first image (See Section 3.2, Page ).

This defines the depth of the reconstruction volume for vertical rotation axis data-sets.