What am I doing wrong ? Pixinsight and debayering

[Vox45]

Hello all,

I am having this issue where I import FITS files in Pixinsight and on debayering them I get funny colors like this:

If I use the same FITS files in DeepSkyStacker I get this ... 

So I started to remove steps from the routine and when I removed the Master Flat from the "Image Calibration" step I got this (single light frame) ... 

so it looks like this part of the process screw things up and I have no idea why ... Anyone with Pixinsight experience here ? I have a hard time understandig why flats would change the color of the debayered light frame... 

 

Thanks!

[scitmon]

Are you actually using the Debayer process or assuming that ImageCallibration does it?

[Vox45]

no I am running the debayer process like this

I debayerd my master flat to see what it looked like ... 

My "problem" might be related to this thread I found on Cloudy nights 

https://www.cloudynights.com/topic/667524-preparing-color-balanced-osc-flats-in-pixinsight/

I will try and see what happens  

 

 

[Aramcheck]

I could be wrong (being very new to Pixinsight myself) but check whether you should be specifying "RGGB". If you open one of your Lights & go to the Process Console on the left hand side of the screen you should see what the CFA pattern is. CR2 files from my Canon are "GBRG" & if I debayer with a diferent setting, like "RGGB" I get a very different colour background.

Worth checking your Master Bias & Darks too I think, as the top image looks like it has more noise than the single frame. (Open the master bias/dark and compare with a single bias/dark frame - the master should have less noise & a smoother appearance).

Cheers
Ivor

[wimvb]

Your master flat is not neutral, which means that it introduces the opposite colour to your light frames. Eg, if the flat frame has RGB values 0.25, 0.5, 0.5 (scale 0 .. 1), the calibration routine will divide:

(Light_R - Dark_R)/0.25

(Light_G - Dark_G)/0.5

(Light_B - Dark_B)/0.5

This will give a higher Red signal in the calibrated light frame than Green and Blue.

Things to consider

• Try to create neutral flats. I believe that EL panels can result in non neutral flats
• Make sure you do the calibration before deBayering (ie, calibraton frames should not be deBayered, only lights just before star alignment)
• Create proper darks for flat calibration

[Vox45]

On 17/12/2019 at 19:29, Aramcheck said:

I could be wrong (being very new to Pixinsight myself) but check whether you should be specifying "RGGB". If you open one of your Lights & go to the Process Console on the left hand side of the screen you should see what the CFA pattern is. CR2 files from my Canon are "GBRG" & if I debayer with a diferent setting, like "RGGB" I get a very different colour background.

Worth checking your Master Bias & Darks too I think, as the top image looks like it has more noise than the single frame. (Open the master bias/dark and compare with a single bias/dark frame - the master should have less noise & a smoother appearance).

Cheers
Ivor

Yes I did check to see if if my Canon produced "RGGB" patterns. It is the case.

 

9 hours ago, wimvb said:

Your master flat is not neutral, which means that it introduces the opposite colour to your light frames. Eg, if the flat frame has RGB values 0.25, 0.5, 0.5 (scale 0 .. 1), the calibration routine will divide:

(Light_R - Dark_R)/0.25

(Light_G - Dark_G)/0.5

(Light_B - Dark_B)/0.5

This will give a higher Red signal in the calibrated light frame than Green and Blue.

Things to consider

• Try to create neutral flats. I believe that EL panels can result in non neutral flats
• Make sure you do the calibration before deBayering (ie, calibraton frames should not be deBayered, only lights just before star alignment)
• Create proper darks for flat calibration

So I've applied this method from Cloudy night I posted earlier in this thread :

https://www.cloudynights.com/topic/667524-preparing-color-balanced-osc-flats-in-pixinsight/

It is possible to prepare a balanced CFA master flat in the same way with PixInsight.  Here is the procedure I use:

Prepare the master flat as usual.
Use the SplitCFA process to create images containing the four separated CFA channels.
Use the LinearFit process to equalize the four images.  Use the channel with lowest intensity as the reference, and apply LinearFit to the other three images.  For me, the CFA0 (red) channel has the lowest intensity in twilight sky flats.
Use the MergeCFA process to recombine the four images, and save the resulting CFA image as a file.  I name the resulting image file "grey-master-flat".  Use this color-balanced master flat (without debayering) to calibrate the light subs.

I ended up with a flat file like this:

A single frame calibrated with the master flat/bias/dark now looks like this (debayered/strech for the purpose of showing the result, I do not do this before registering and stacking)

once star aligned I get this (!!!)

and after stacking I get this final image:

All those vertical bands ! and the dark corners on the upper and lower right side.

I am doing some tests right now, with different combinations of duration/light intensity and with/without the CLS-CCD filter and I am getting results all over the place. I am obviously doing something wrong, I will need to dig a little deeper and read again the "Inside pixinsight" book

Edited December 19, 2019 by Vox45

[wimvb]

If you use 10 s exposures with your Canon, I think your subs are underexposed. The lines/bands are a typical read pattern of a dslr sensor.

Flats should be as neutral as possible, but PI can handle a colour cast caused by the calibration process. Post processing will take care of this. It will be corrected during backgound neutralization. Personally I would try to avoid doing any processing on calibration frames, other than calibration and integration.

[Vox45]

1 hour ago, wimvb said:

If you use 10 s exposures with your Canon, I think your subs are underexposed. The lines/bands are a typical read pattern of a dslr sensor.

Flats should be as neutral as possible, but PI can handle a colour cast caused by the calibration process. Post processing will take care of this. It will be corrected during backgound neutralization. Personally I would try to avoid doing any processing on calibration frames, other than calibration and integration.

It took me a while to understand where you got the 10s exposures number from  I see that you got that number from the name of the file in the debayer window I posted. Actually, it was 300s subs and I took a couple of 10s and 15s to help with the core of M42. The rest are 300s subs.

Edited December 20, 2019 by Vox45

[wimvb]

How and when in the process did you mix the short an long exposures? 

[Vox45]

6 hours ago, wimvb said:

How and when in the process did you mix the short an long exposures? 

I have to admit that I just threw everything into the mix. Your comment gave me pause ... So I did the whole process once more but I only used the 300s light frames this time.

this is the result after integration of 12x 300s light frames:

no more vertical bands ! but I still get the dark corners on the upper and lower right side.

Compared to 12x 300s + 10x 10s + 10x 15s + 10x 20s giving this result :

So it seems that throwing all the files with different duration into the mix was not such a good idea  I was biased (no pun intended) by the fact that I just throw everything in DSS and get good results... 

[wimvb]

Pixinsight has a process/script called hdr composition. Try that. Otherwise the integration process just calculates the average of all the subs.

Btw, the vertical lines can be removed with CBR, canon banding reduction.

• Fastrotate 90 deg
• CBR
• Fastrotate back

[Vox45]

Thank you for you advice  I will try that !