Previewing

Previewing#

Previewing crops, or slices, the input data. It can remove unused regions and reduce processing time, particularly while sweeping parameters. See Enabling data preview to start configuring it.

Previewing in the loader#

The Standard tomography loader provides a preview parameter for selecting part of the input data.

Note

HTTomo assumes a three-dimensional array whose axes are the angular dimension, the vertical detector (\(Y\)) and the horizontal detector (\(X\)), in that order (see Fig. 7).

3D data

Fig. 7 3D projection data and their axes#

The preview parameter#

The preview parameter has one field per axis. Each field accepts start and stop values:

preview:
 angles:
   start:
   stop:
 detector_y:
   start:
   stop:
 detector_x:
   start:
   stop:

The stop value is excluded, as in a Python slice. For example, the following selection loads projections 20 through 99:

preview:
  angles:
    start: 20
    stop: 100

Note

continuous_scan_subset also selects the angular range. When it is set, it replaces preview.angles. The command-line --continuous-scan-subset option takes precedence over both values. See Continuous-scan subsets.

Using the full dataset#

Omitting preview selects the full dataset without cropping.

Enabling data preview#

Crop either or both detector dimensions to reduce the data size and accelerate processing.

Note

Removing blank detector regions reduces the reconstructed volume and can also accelerate post-processing.

The following projections show vertical and horizontal cropping.

Before cropping pic1 and after pic2

  1. Crop blank regions from the top and bottom of the vertical detector (\(Y\)), as shown in Fig. 8. Inspect the raw projections to identify regions that remain blank throughout the scan.

    preview:
  detector_y:
    start: 200
    stop: 1800

    This selects slices 200 to 1799, producing a vertical dimension of 1600 pixels. The equivalent Python slice is [:, 200:1800, :].

3D data, Y slicing

Fig. 8 Cropping detector- \(Y\) dimension of 3D projection data#

  1. Crop blank regions from the left and right of the horizontal detector (\(X\)), as shown in Fig. 9.

    Warning

    Horizontal cropping can disrupt automatic centering and introduce reconstruction artefacts, particularly with iterative methods. Crop the \(X\) dimension conservatively.

    preview:
  detector_x:
    start: 100
    stop: 2000

    The equivalent Python slice is [:, :, 100:2000].

3D data, X slicing

Fig. 9 Cropping detector- \(X\) dimension of 3D projection data#

Combine both operations as follows:

preview:
  detector_y:
    start: 200
    stop: 1800
  detector_x:
    start: 100
    stop: 2000

Using begin, mid and end with offsets#

Use begin, mid and end instead of absolute indices when the input dimensions are unknown. They may be used in the angular range as well as the detector ranges. Adjust them with start_offset and stop_offset:

preview:
  detector_x:
    start: begin
    start_offset: 100
    stop: end
    stop_offset: -100
  detector_y:
    start: mid
    start_offset: -50
    stop: mid
    stop_offset: 50

This removes 100 pixels from each end of detector_x, equivalent to [100:-100], and selects 100 pixels centred on detector_y.

Note

begin, mid and end identify the first, middle and last indices of a dimension, respectively.

Using mid by itself#

The detector_y and detector_x fields also accept mid without start or stop:

preview:
  detector_y:
    mid

This selects the middle three slices of the specified dimension.

Warning

The angles field does not support mid.

Omitting preview fields#

You may omit unused dimension fields and start or stop values.

Omitting one or more dimension fields#

An omitted or empty dimension field selects that entire dimension. The following configuration therefore selects the full dataset:

preview:
  angles:
  detector_y:
  detector_x:

Omitting the start or stop fields#

For each dimension:

  • Omitting start begins at index 0.

  • Omitting stop continues to the end of the dimension.