Monochrome flats with color CMOS camera

bemo47

Member
Dec 16, 2016
10
0
Hi,
I use an ASI1600MC cmos color camera.
But when i the my flats i get a low red channel and two high blue and green channels, totally unbalanced.
What is the best way to solve that ?
Converting my color flats in monochrome flats and how ?
Balancing RBD channels of my flats before using them and how ?
Thanks for your advices,
Bernard
 

fmeschia

Active member
Sep 19, 2017
26
0
I run this PixelMath expression on my master flat to make it neutral gray. Be aware that it reduces its effective resolution by a factor of 2, so the flat won’t be so effective for correcting very small-scale (1 pixel) patterns:

0.25*(pixel($T,floor(x()/2)*2,floor(y()/2)*2)+pixel($T,floor(x()/2)*2+1,floor(y()/2)*2)+pixel($T,floor(x()/2)*2,floor(y()/2)*2+1)+pixel($T,floor(x()/2)*2+1,floor(y()/2)*2+1))

Francesco
 

bemo47

Member
Dec 16, 2016
10
0
whao, thanks for the advice, but i don't know pixelmath, i never used that. So i did the test and i got a perfect alignment of all 3 colors, perfect gray. That's exactly what i wanted.
why do we have this issue ? is it linked to CMOS cameras ? or only some sensors like mine ?
 

pfile

PTeam Member
Nov 23, 2009
5,099
41
mainly what's important is to make sure the SNR of the weak channel of the flat is high enough to not mess up the SNR of your subs. any color imbalance that the flat imparts to the sub will be undone when you color calibrate the image. whether or not your flats gave you a color cast, you have to do this step anyway.

when i used a DSLR and a CLS filter i used to make T-shirt flats with a pink T-shirt. that would boost the red channel enough that the flats were nearly grey to start with.
 

dld

Well-known member
Sep 24, 2017
147
9
Also, the strongest channel should remain away from the non-linear part of the sensor for the given gain and offset. If you intentionally overexpose a flat frame, you'll notice that all channels will collapse at a strong peak towards the right part of the histogram (the one obtained by HistogramTransformation). If you're using an exposure which is high enough to for the lowest channel to have high SNR and low enough for the highest channel to be away from being saturated, you will be fine. Color balance will be corrected at later steps by PhotometricColorCalibration.
 

bemo47

Member
Dec 16, 2016
10
0
not sure to understand all u said above, but here below are example of the unbalanced flats and master flats, before and after the pixelmaths applied, so clearly after pixelmaths its very well balanced.

also i don't know how to analyse this balance in my images, the best i found is to show the histogram transfo before any changes, here we can see that there is a discrepancy between two set of colors, but we can't know which color.... is there a better way to analyse linear color images, or non linear... ?


here is the balance i get in my master flat and photometric calibration doesn't correct the red dominant at the end of the treament.
1585562446865.png

and here the same master after pixmaths as proposed
1585562542268.png
 

fmeschia

Active member
Sep 19, 2017
26
0
not sure to understand all u said above, but here below are example of the unbalanced flats and master flats, before and after the pixelmaths applied, so clearly after pixelmaths its very well balanced.

also i don't know how to analyse this balance in my images, the best i found is to show the histogram transfo before any changes, here we can see that there is a discrepancy between two set of colors, but we can't know which color.... is there a better way to analyse linear color images, or non linear... ?


here is the balance i get in my master flat and photometric calibration doesn't correct the red dominant at the end of the treament.
View attachment 8081

and here the same master after pixmaths as proposed
View attachment 8082
The problem I see is that at least one of your channels in the Bayer matrix is quite underexposed. Averaging the channels may help a bit, but the real “cure” would be to change the color of the light source you use, to have less red.
 

fmeschia

Active member
Sep 19, 2017
26
0
Also what other people are saying is that it doesn’t matter what is the color of your flat, as long as the strongest channel is still in the linear part of the curve, and the weakest channel has sufficient SNR. Any color unbalance will be corrected during color calibration anyway.
 

pfile

PTeam Member
Nov 23, 2009
5,099
41
you can click on the R, G, B icons (next to RGB/K) to see just the one channel. @fmeschia i think he said the flats have a strong blue cast so he needs to increase the red light. if the red is low in the flat then it will be emphasized in the calibrated result (division by a smaller number).

anyway @bemo47 your background tolerance probably needs to be adjusted in PCC. it should be able to fix the color cast. hover your mouse over the background and see what the highest channel's pixel values are like and then set the background upper limit to something just above that value.
 

fmeschia

Active member
Sep 19, 2017
26
0
Hi Rob, you’re right about the color cast, sorry. I think @bemo47 can’t see the individual channels in HT because he’s looking at a non-debayered master flat.
 

bemo47

Member
Dec 16, 2016
10
0
Hi Rob, you’re right about the color cast, sorry. I think @bemo47 can’t see the individual channels in HT because he’s looking at a non-debayered master flat.
whao, every day i will learn something.... astrophoto is a very long way for me sorry :)
thanks to both of u, and yes, when i debayer my flat i can see the spread of the 3 colors easily !!
but then.... with these color levels, may be the red is really to low to give a good flat and the ability to correct the color levels later with PCC... may be i can test what @pfile said about checking the background with PCC
but also i will make flats on the sky to check if i have the same difference in colors.
i hope its not due to my flat panel, i boutgh a costly one, Aurora, in order to avoid any problems regarding flats.... :mad:
 

Attachments

pfile

PTeam Member
Nov 23, 2009
5,099
41
yes, yes.... i knew that.... but.... :confused::confused:
haha, thks anyway for ur patience !
to be continued soon :)
sorry i was meaning to reply to francesco with that. i saw the multiple humps and misinterpreted what i was seeing. the different channels' histograms should appear in color in that view when you're looking at the histogram of and RGB image.
 

bemo47

Member
Dec 16, 2016
10
0
hi, some news, after many trials, i got to the conclusion that i can't use TheSkyx for flats as it doesn't give the ability to manage the white balance when we take images, and my flats have always this big difference betxween red and blue/green.
So i will use Sharpcap, with this tool i can manage the white balance and get a better balance between colors.

here the flat taken with TheSkyx without white balance modified
1585645231158.png

here the flat taken with sharpcap with modified white balance, hope it will be better as flat !
1585646466643.png
 
Last edited:

pfile

PTeam Member
Nov 23, 2009
5,099
41
you really don't need to worry about the white balance, and you shouldn't modify the flats. it doesn't do anything to multiply one channel by some fixed amount. the underlying data is still exactly the same. all you need to worry about is that you have enough flat subs that the SNR of the weakest channel is acceptable.
 

pfile

PTeam Member
Nov 23, 2009
5,099
41
one strategy is the "million electron flat" - find the ADU of the dimmest part of the weakest channel in one subexposure, convert that to electrons using the gain setting of your camera, and then the number of subs you want is (1,000,000 / electrons in weakest part)

rob