White balance and debayer interpolation

Given a DSLR (Canon 6d) and wanting to calibrate my subs (so no debayer on conversion to FITS) it seem that there’s no way to white balance before debayer interpolation. Interpolation algorithms are generally non-linear, so are supposed to use white balanced inputs otherwise you get colour artefacts in high contrast regions. How am I supposed to do this correctly using Siril? Other AP tools (and RAW development tools) all support white balancing before interpolation. But not Siril?

Astrophotography software never do white balance before preprocessing. This is always done on the stacked image, with the color calibration tool. Based upon stars.

I know Siril doesn’t, but plenty of others do exactly that (allow colour balancing before interpolation). My point however was that most debayer interpolator algorithms expect linear white balanced input, so colour balancing after interpolation is suboptimal. I have no problem with AP software leaving the before/after option to the user - I’m just surprised that Siril (alone as far as I can see) doesn’t allow you to feed the interpolation algorithms with the data they’re designed to consume.

Ok, but for daylight images. Pixinsight, APP don’t do this.
So no, Siril is not alone. Astrophotography image processing is quite different from daylight images.

One of our next feature will be CFA debayer where interpolation is not needed anymore.

You’ve still not addressed my point that you offer interpolation algorithms (e.g., RCD) which rely on colour balance inputs, but don’t provide any means of colour balancing before interpolation. You seem to be defending the Siril design as a ‘matter of principle’ (“Astrophotography image processing is quite different…”) rather than because it’s technically correct (which it isn’t), which is a pity. I understand that for many AP users colour is an aesthetic decision, but my camera is well calibrated for colour (I know the RGB multipliers and profile to get accurate star colours) so it makes sense to use those at the start of the workflow and allow the interpolators to do their job as accurately (minimum artefacts) as possible.
I look forward to seeing your non-interpolating debayer (drizzle?).

For the avoidance of doubt…

  1. Apart from my comment/complaint, I think Siril is excellent.
  2. Having upgraded to 1.2.1, I now have an (albeit somewhat laborious) work-around (CFA_split → PixelMath → CFA_merge).
    Thank you.

The next version (1.4.0) will enable a color conversion matrix to be applied to the entire sequence. However, this will not be part of the standard workflow.

But here again, this is a false problem: the problems of daytime images are very different from those of astrophotography images (apart from the brightest stars, there are no high-contrast areas). And even here, artifacts are well managed with the RCD algorithm). That’s why there’s no astrophotography software that does this.