Ignore masked detectors in smooting

[Owen Arnold]

Nearest neighbours will ignore monitors, but it also needs to ignore masked detectors.

This can be acheived by.

• Ensuring that the masking algorithm clears the nearest neighbours map on the workspace.
• Have the nearest neighbours map ignore masked detectors when it builds the map.

Note that when using the radius method of smooth neighbours. Users will end up with less neighbours than they used to have when this is implemented (if their are masked detectors). Using the 'Exact' method of selecting neighbours they will end up with the same number since the NearestNeighbours should keep looking for neighbours until it meets the specified value (while still ignoring masked ones). Pascal also wants the option to override this if possible, so that he can select 8 nearest neighbours, but say only end up using 7 if one is masked.

[Owen Arnold]

Having seen this issue with Pascal. I'm upgrading it to critical, smoothing will smudge the results of spurious detector results, and the converstion to QSpace may result in quite significant regions of the visualisation being affected.

[Owen Arnold]

refs #4150. Fix slicing issue by introducing v-destructor.

Changeset: 447885ab4768d17ab289834821a148f9c319c3cd

[Owen Arnold]

refs #4150. Use reset instead of swap

Changeset: ffb40c86ed54c12837f357ec6f141535ea19148b

[Owen Arnold]

refs #4150 Fix identified warnings

Changeset: d451b23c1c964597a871d3ae24d2e60f6b0e522b

[Owen Arnold]

It's now possible for an algorithm to clear the masking on a MatrixWorkspace. Creational methods just introduced should also help with necessary nn map rebuilds as creation is abstracted.

[Owen Arnold]

refs #4150 SmoothNeighbours updated.

Can now speicify that SmoothNeighbours ignores or includes masked detectors. Appears to work from instrument view. This is an intermediate step. Next thing to do will be to include union criteria for limiting searching by radius as well as by number of neighbours. Changeset: 6f75115bcc0c00ab4dfe53e2447624757377ae9f

[Owen Arnold]

refs #4150 Fix Type usage warning

Changeset: eed8e718856b9643743d8f40394f015495e4a235

[Owen Arnold]

refs #4203 #4150. Fix iterator usage error + warnings

Changeset: 3233895b75dcbb9fbfc1af2853ccb0e1e974e84b

[Owen Arnold]

refs #4203 #4150. Fix unused arg warning

Changeset: ea2a83db6c14bc1d2d26eed49755e46ebf7df3da

[Owen Arnold]

refs #4203 #4150. Minor change to wiki docs

Changeset: 2b6fc8367a1741d54f1dab584e617a5a1be0b16c

[Owen Arnold]

refs #4203 #4150. Minor change to wiki docs

Changeset: 020e92fea6d85318dc5b0ae3a633a6228d9a0f49

[Owen Arnold]

Tester: You can test this as follows:

-Open up a noisy workspace in the instrument view. One of the WISH raw files from the data directory would be fine.

-Mask a rectangular region of this workspace using the tools in the instrument view.

-Run SmoothNeighbours with RadiusUnits=NumberOfPixels Radius=200 and Number of neighbours = 46 and ensure that IgnoreMasks=True on this workspace. This should ensure a large number of neighbours are including in the moving smoothing.

-Open the output workspace in the instrument view and it should be obvious that the masked region has not been smoothed over. It should look much noiser that the surrounding smoothed regions.

[Peter Parker]

Ran SmoothNeighbours on a (large) WISH file from the archive - the difference between the noisy and noise-cancelled versions was clear, and the masked detectors were unaffected.

[Stuart Campbell]

This ticket has been transferred to github issue ​4997