include scaling function for post-processing

[mrheinnecker]

I discovered the following problem at the data post-processing of segmentations:

I will give a very precise example but i think the feature could be of advantage for other things as well.

To achieve segmentations of cell boundaries it made sense (at least for some cells) to do a segmentation of the background and use the negative to get the cell area.

As the background is segmented only in the mask area, the segmentation gives only the background area inside the mask. That means that on the full image I get the segmented area (>0) as well as the area of the cell (will be zero) but also the background area outside the mask will be zero.

To get over this, an overlay of the mask and the segmented background area can be used zo extract only the zeros of the cell.

The problem here is:

For datasets with non-isotropic input, cebra-em can perform a scaling and interpolates the "missing" slides for the input image and the mask. But these interpolations are not stored somewhere due to their size.

That brings up the problem when I want to overlay the mask and segmentations, the z-resolution differs and not every layer of the segmenattion is covered by the mask. As I dont want to lose the interpolations in the segmenattion, I need to scale the mask manually.

Of course i could implement this by myself, but the functionality is somewhere inside of the cebraEM functions and it would probably make sense to have it as a function exported to the namespace?

For manual corrections afterwards this would maybe also make sense for the raw EM image, which would require the same functionality.