Adds methods to display surfels/quad with normal vectors

[dcoeurjo]

• Ok for both Viewer3D and Board3D.

• usage should be similar to

   Display3DFactory<>::drawSurfelWithNormal(viewer,
                                         surfel,
                                         normalVector);
  

• surfel being a Khalimsky cell (signed or not) and normalVector a unitary normal vector.

Illustration with a 100^3 digital sphere surface with "exact" normal vectors:

TODO:

• assert that the kcell/cell is a 2-cell

BE CAREFUL : with QGLVIEWER, you need to press "D" to enable double-face rendering

[kerautret]

Nice! I look it ;)!

[kerautret]

By checking the Board3D I don't see the normal both with blender or Meshlab you use another viewer ?

[JacquesOlivierLachaud]

There are two versions of Display3DFactory::drawSurfelWithNormals, one for Cell, one for SCell. In fact, both types could be identical (CellularGridSpaceND implementation dependent). Hence I suggest to rename the methods

• for oriented n-1-cell (SCell), we can leave drawSurfelWithNormals since a surfel is an oriented element.
• for unoriented n-1-cell (Cell), we could name it drawCellWithNormals or drawUnorientedSurfelWithNormal.

It would also disambiguate the boolean (double side) for normals. A Cell is double sided, a SCell no.
What do you think ?

[JacquesOlivierLachaud]

Apparently embedK exists only for Cell and not for SCell.
The function is thus not instantiable.

[dcoeurjo]

agree.. there is a embedSK(k) for that.

[dcoeurjo]

oups it's embedKS(k). Fixed

[dcoeurjo]

wouldn't drawCellWithNormals be confusing with the fact that it does not consider any cell but only (n-1)-one ?
What about drawOrientedSurfelWith... and drawUnorientedSurfelWith... ? (a bit long btw)

[JacquesOlivierLachaud]

Go for drawOrientedSurfelWith... and drawUnorientedSurfelWith
It is fine with me.

[JacquesOlivierLachaud]

I still have some displaying issue when the estimated normal has a negative scalar product with the trivial normal. I must check a little bit more (are you "correcting" the normal at display time ?).

[JacquesOlivierLachaud]

Ok. I looked in DGtal::Display3D< Space ,KSpace >::addQuadWithNormal and normals are reversed if the scalar product is negative. I am not sure this is always desirable. Look at the example below. The estimated normal is close to the z-axis, so for some surfels (containing the z-axis) they appear black.
A hack is projecting the normal onto the surfel to limit this problem.
A cleaner method is to let the normals as is (with some drawbacks when normals are not correctly oriented).

[JacquesOlivierLachaud]

[dcoeurjo]

In your example, do you use the "double-quad" feature ?

[JacquesOlivierLachaud]

No. When I enable it (key d), it is much worse (see image below). I suspect this is due to a too small epsilon for the precision of the depth buffer in my graphics card.

[dcoeurjo]

@JacquesOlivierLachaud I've changed couple of things (function renaming, reorientation disabled for signed cell, ...)

[JacquesOlivierLachaud]

I've got still (more) trouble when displaying surfels with drawOrientedSurfelWithNormal.

The result is identical with or without enableDoubleFace. Only surfels with trivial normals aligned with Y are correct.

[JacquesOlivierLachaud]

I have corrected a few things and modified Display3D. I think it is fine now. Merging.