debayer confusion

[alacant]

Hi everyoneç

Siril 0.9.9 Ubuntu 18.04

I'm migrating frrom DSS to Siril and am not sure when to debayer. I have a 700d. Let's say I have bias, flat and dithered light frames.

My DSS workflow is decided for me: I load all the frames into DSS and hit register, then stack. It builds master bias, master flat (after first removing the master bias from each flat) then subtracts the bias from the light frames, finally dividing the lights by the master flat.

I'm not sure when it debayers.

In Siril I have to do this manually. I have tried debayering all bias, flat and light frames, which seems to work fine.

Question: What's the correct workflow? Which frames do I debayer and when?

TIA and clear skies

[vlaiv]

Hi,

Debayering can be done last, on light subs only, after calibration.

Before that, you can treat all subs as mono, just do regular calibration like you are doing.

 

[han59]

I think it is better to start with raw images and use them to stack and let the stacking program to apply the Bayer matrix rather the de-bayer. I don't know how SIril it is doing but that is the purest method.

For this I use the program DCRAW with parameters -D -4.   ( Canon EOS 700D is suported) Then you get 16 bit PGM files which look mono and are UN-BAYERED. Those can be converted to 16 bit FITS,  at least that's what I'm doing in the ASTAP stacking program.

I don't know how DSS or Siril are internally working but the process should be as follows:

The master flat is calculated as follows

masterflat: = (1/n ∑ [flat] - 1/n ∑ [flat darks] )

Where a bias image could be used as flat-dark image. The masterflat should be averaged by a 2x2mean to remove Bayer matrix artifacts.

Each image is calculated as:

 (image- {∑ [darks]/n} ) / master flat

Then the Bayer matrix is applied and finally the images are stacked in mode average or sigma clipped.

So for average stack:

final image:= 1/n ∑ Bayer(image)

So again using un-bayered raws is better then to de-Bayer coloured images.

Han

 

 

 

 

 

[alacant]

11 hours ago, han59 said:

I don't know how SIril it is doing

This is probably totally wrong but at least it produces a coloured fits file at the end!

Removing the Siril specific stuff, I think it's this:

stack bias -> master bias

subtract master bias from each flat

stack flats -> master flat

subtract master bias from each light

divide each light by master flat and then debayer 

star align the frames (use say the green layer)

stack the light frames -> stretchable fits in colour

Thanks for your time and clear skies

 

 

 

 

 

[han59]

11 minutes ago, alacant said:

stack bias -> master bias

subtract master bias from each flat

stack flats -> master flat

subtract master bias from each light

divide each light by master flat and then debayer 

star align the frames (use say the green layer)

stack the light frames -> stretchable fits in colour

 

Assuming you start with coloured images and don't have darks, the above looks good to me if you remove the strikethroughs. Why debayer? The red,green and blue have the same pixel position so can be handeled/processed combined.

[alacant]

1 hour ago, han59 said:

Why debayer?

If I don't, I get only a one layer monochrome image with a grid pattern. To get rgb, I have to debayer. But please tell me otherwise:) 

[vlaiv]

3 hours ago, alacant said:

This is probably totally wrong but at least it produces a coloured fits file at the end!

Removing the Siril specific stuff, I think it's this:

stack bias -> master bias

subtract master bias from each flat

stack flats -> master flat

subtract master bias from each light

divide each light by master flat and then debayer 

star align the frames (use say the green layer)

stack the light frames -> stretchable fits in colour

Thanks for your time and clear skies

 

 

 

 

 

Not wrong at all, that should work perfectly - exactly like I've said before, you can do your full calibration without debayer (and you should).

When you have calibrated frames, then you debayer and proceed to stacking. You can split frames to get R, G and B channel and stack each to their own stack, or you can create "color" images

and stack those. I prefer first method - three separate stack, at the end I align them and do RGB combination to get final result.