Smoothing application enhancement

Summary

Some modifications of the Smoothing application. In particular the Gaussian mode now has two additional parameters : type.gaussian.maxerror and type.gaussian.maxwidth. These parameters were available in the ITK filter, but were not exposed in the application

Closes #1961 (closed)

Implementation Details

Applications

Smoothing :

• new parameter type.gaussian.maxerror : The algorithm will size the discrete kernel so that the error resulting from truncation of the kernel is no greater than type.gaussian.maxerror
• new parameter type.gaussian.maxwidth : Set the kernel to be no wider than maxwidth pixels, even if type.gaussian.maxerror demands it.
• the radius parameter has been renamed into stdev (because this is what it is, see the TODO in code)
• Some modernization (auto, RegisterPipeline())

FastNLMeans :

• Some modernization (auto, RegisterPipeline())

Tests

The tests for the Smoothing application have been updated:

• New test for the Gaussian mode (there was no test for this mode)
• New test for the Anisotropic mode (there was no test for this mode). Note that I needed to set a tolerance value for this test (1e-6), because the result were different on windows 8 (x86). The differences are probably caused by numerical error, and we can't do much about it anyway because we are using an ITK filter.
• The apTvUtSmoothingTest_InXML now depends on the apTvUtSmoothingTest_OutXML test, instead of using a stored baseline.
• The apTvUtSmoothingTest has been removed, there is no need to test 3 times the mean mode with the same parameters. (see InXML and OUTXML tests).

Copyright

The copyright owner is CNES and has signed the ORFEO ToolBox Contributor License Agreement.

---

Check before merging:

• All discussions are resolved
• At least 2 votes from core developers, no vote.
• The feature branch is (reasonably) up-to-date with the base branch
• Dashboard is green
• Copyright owner has signed the ORFEO ToolBox Contributor License Agreement
• Optionally, run git diff develop... -U0 --no-color | clang-format-diff.py -p1 -i on latest changes and commit

changed milestone to %7.1.0

added feature refactoring labels

assigned to @ctraizet

added 1 commit

• 444569e3 - DOC: fix error in documentation

Compare with previous version

changed the description

added 1 deleted label

added 1 commit

• 877e4898 - BUG: add tolerance on the anisotropic smoothing test

Compare with previous version

changed the description

Works fine on my superbuild (Ubuntu 18.04). I tried to find some references on the algorithm, because it's not so simple to understand the parameters. In my example (extract of a PHR image), a greater maxerror (0.5 instead of 0.01) produced to a sharper image... Maybe the doc should reference ITK's doc for this kind of filters.

@ytanguy I added some reference to the ITK filters and to the papers of the algorithm

I also added reference to the new NL means application (and reference to Smoothing in NL means)

added 1 commit

• 03495bd7 - DOC: add references to the paper of the algorithms and to the ITK filters

Compare with previous version

added 67 commits

• 03495bd7...a9340fdf - 60 commits from branch develop
• 89ab63f2 - DOC: rename mode.gaussian.radius parameter into mode.gaussian.stdev
• 07b601e9 - ENH: added parameters for maximum error and maximum kernel width in Smoothing
• 566a195b - ENH: updated test for the Smoothing application
• d47e9321 - ENH: some code modernization
• 7d27b411 - DOC: fix error in documentation
• 893bff1a - BUG: add tolerance on the anisotropic smoothing test
• c0f61f42 - DOC: add references to the paper of the algorithms and to the ITK filters

Compare with previous version

mentioned in commit f6d4e7bb

merged

Ok thank you @ctraizet !

mentioned in issue #2036 (closed)