Normalization of mapping tables

[TjarkMiener]

The normalization in image mapper isn't working properly. The core problem is the following:

1. Firstly each mapping table is normalized by normalization(self, mapping_matrix3d, num_pixels). For example for the mapping table of the one camera pixel in the figure, the weights of the output grid points (red and green and from more Delaunay triangulation around the pixel) are normalized.
2. Then, the mask is applied by mask_interpolation(self, mapping_matrix3d, nn_index, num_pixels, pad). The mask will set the red grid points to zero and keep the green points. This will somehow destroy the normalization and we will end up having less intensity/energy in the transformed image.

Applying the mask before the normalization will result in the same problem.

[TjarkMiener]

I added an optional argument in image_mapper for mask interpolation in commit e1639c1

[aribrill]

When the argument for mask interpolation is set to True, does this issue still persist? If so, I think this issue should be reopened. If I understand correctly, the point of mask interpolation is to prevent contamination from zero padding which causes a border artifact, so it should be set to True by default.

[TjarkMiener]

Yes! When mask interpolation is True it's still persist. However, I think this is an issue that occur by construction and can't be avoided.