Non-negative fluorescence output?

[ziw-liu]

Currently waveorder can output negative fluorescence intensity. This leads to inconveniences in visualization (auto-contrast will boost background noise) and downstream numeric processing. I can think of 2 ways to guarantee a non-negative output:

1. Clip by 0.
2. Shift the histogram to the right.

There could be a flag passed to 'apply inverse' to enable this behavior.

[talonchandler]

I think it sense to have clipping as a default-off option clip=False. My rough experience is that downstream consumers (people and algorithms) often expect smooth reconstructions on the scale of a few pixels, and clipping like this creates sharp edges that can create problems downstream.

Shifting the histogram avoids the sharp edges, but I think it will lead to different kinds of confusion---should the maximum value shift based on distant regions in the image?

[mattersoflight]

@talonchandler @ziw-liu I agree that clipping should be avoided to avoid texture or high-spatial frequency artifacts. I would add a constant offset, just like we adjust the black level for a camera. It can be a parameter to apply inverse and we can default it to a small value that avoids negative output in most cases.