Fix issue in DGtal PR#1086

[jlevallois]

It fixes the issue with testIntegralInvariantVolumeEstimator in your PR DGtal-team#1086

• I add a check in DigitalSurfaceConvolver to avoid creating cells outside the domain (with non periodic spaces, it was always returning 0).
• I also add 4 new functions in KhalimskySpaceND to check if a cell is valid or inside with only is Khalimsky coordinates (no need to create the cell).

Is it ok for you ?

[rolanddenis]

Thanks for doc fix ! Too many copy-paste ;)

[rolanddenis]

Thanks for the fix ! Some remarks but I can make the modifications by myself if you prefer...

[jlevallois]

I have a weird bug with your PR (even without my fixes if I remove some asserts on sSetKCoords())

If I compute the (mean) curvature with 3dCurvatureViewer (DGtalTools) using results of previous computation for all surfels (aka with shifting masks), I get this:

If I compute the (mean) curvature with the full kernel each time, I get the good result:

It works fine with the master branch of DGtal.
Do you have an hint about why I get this error?

Full command:
./3dCurvatureViewer -i bunny64.vol -r 15 -u 255

[jlevallois]

Ok, it's not exactly the same results than with master branch:

So, maybe it's due to the shifting with innerSpel outerSpel (with sDirectIncident() and sIndirectIncident() ) ?

[rolanddenis]

Hum 😱

[rolanddenis]

Along which axis do you have those oscillations ? X ?

[rolanddenis]

I merge the PR in order to continue the discussion on the DGtal-team#1086 .