Versions Compared

Version Old Version 29 New Version 30
Changes made by

Kaz Wanelik (Unlicensed)

Kaz Wanelik (Unlicensed)

Saved on

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

...

VoCentering
Process categoryBrief description

Computational demand

for typical tomography data

(low, medium, high)

Comment(s)Reference(s)Common alternative process(es)
filter

To find an optimal value of CoR automatically.

High

Since the computational cost of this auto-centring process is high, it is normally applied to a small, representative subset of data (this can most conveniently be done with the aid of the preview parameter of this process).

  
  • Expects to receive image data of rank 3.

    •  

    		Reliable method for calculating the center of rotation in parallel-beam tomography VoCenteringIterative

    ...

    ItemParameter nameParameter formatExample(s)Comment(s)
    Parameter valueEffect
    1

    preview




    1. Note that the preview parameter has a nested-cumulative behaviour, i. e. if one specifies VoCentering's preview parameter together with NxtomoLoader's preview parameter, then VoCentering will effectively be selecting a subset from NxtomoLoader's own subset (rather than the entire dataset). Note also that VoCentering's subset must always be specified by indexing NxtomoLoader's own subset (rather than the entire dataset). For instance, one must set VoCentering's preview parameter to [:, 0, :] in order to select the initial slice of any dataset previously loaded by NxtomoLoader, which may, for example, be the [:, 123:456, :] subset of the entire dataset.
    2. If you use any kind of parameter tuning before applying VoCentering, then the latter will automatically receive (from the immediately preceding process) a dataset of order higher than 3, in which case you must slice the incoming dataset by specifying the preview parameter of VoCentering in such a way as to generate . In this case, if you leave VoCentering's preview parameter at its default value (i.e. [ ]), then VoCentering will process the entire rank-4 data, which is hardly ever desired. To avoid this situation, you must set VoCentering's preview parameter to be a desired rank-3 sub-set slice of that higherincoming rank-rank 4 dataset. For instance, if you have generated a new dataset of shape (D, img_W, img_L, T) by applying a single parameter-tuning operation to subjecting the original (D, img_W, img_L) data to a single parameter-tuning operation, then this your new, rank-4 dataset needs to be sliced in the last dimension, e. g. if the preview parameter of VoCentering is set to be the [:, mid, :, 0] slice of the that parameter-tuned dataset , then will give you the resulting rank-3 subset selects containing the middle sinogram from (D, img_W, img_L), corresponding to created with the initial value of the your tuning parameter.    
    2

    start_pixel




    		An initial estimate for the pixel coordinate of an optimal CoR.
    3

    search_area




    		If the value of the search_area parameter is set to the default interval of (-50, 50), then VoCentering will attempt to search for an optimal value of CoR in the (start_pixel - 50, start_pixel + 50) interval. For some datasets, this default range of search (100 pixel) may be not sufficiently large to include an optimal value of CoR, leading to subsequent sub-optimal reconstruction. Therefore, if the value of CoR determined by VoCentering is found to coincide with one of the search-interval limits (i. e. either start_pixel - 50 or start_pixel + 50), then this value of CoR may not necessarily be optimal, and one should then re-run VoCentering with a larger value of the search_area parameter to confirm this result.   
    4

    in_datasets





    5search_radius



    6ratio



    7out_datasets



    8

    datasets_to_populate





    9row_drop



    10step


    		Floating-point or integer value in pixel units.

    ...