What to do when Photometric Color Calibration Doesn't...

RodrigoQ

Member
Jul 11, 2020
5
0
Thank you so much for your fast Reply Juan. I'm new using this software and trying to understand a lot of things. I'll try to proceed as you suggested (Using DBE instead of ABE ) Will be a challenge find those free sky regions, the amount of samples affect the process ?? I tried with 7 samples and after perform the DBE and run the PCC i still have the green cast but in a less amount of earlier. Probably Did i a mistake on the DBE?

Thank you so much!!

Rodrigo Quiroga
 

Attachments

wrichards

Well-known member
Jun 22, 2020
69
0
I have the exact same issue. The original stacked image has a strong red tint. After running either ABE or DBE, the red is gone and the color balance looks great. Then after running PCC, the image has a strong green tint to it. It doesn't matter whether I use ABE or DBE, or how many points I use for DBE, how big or small they are, etc.

The first attached image shows the original stack with the DBE points assigned. They have a radius of 12 and the vast majority contain no stars - not even tiny ones. A very few contain single pixel faint stars.

The 2nd image is after DBE - looks very nice.

The 3rd image is after PCC - quite green.
 

Attachments

wrichards

Well-known member
Jun 22, 2020
69
0
I'm pretty sure this isn't a Background Extraction issue at all. I just took my unrevised, stacked image and ran it through PCC without doing any background extract at all and the image still turns green.
 

Attachments

pfile

PTeam Member
Nov 23, 2009
5,622
104
well for the most part, you don't have to change the defaults. however, just like the BackgroundNeutralization tool, PCC needs to sample the background of the image to neutralize it properly. basically you can optionally define a preview over an area of the image that you know to represent the background only, and also set the threshold slider to tell PCC/BN how bright the brightest background pixels are. if your un-color-calibrated image has a huge disparity between brightness across the 3 channels, one of them will have really high signal in the background. if that level exceeds what is configured in the tolerance slider, PCC will ignore it, thinking it is signal and not background. so you should take a look at the background pixel values with the readout tool and set the slider just slightly greater than the level of the brightest background channel. if the field is really busy with stars and nebula it might be a good idea to set the background area of interest as well.
 

RodrigoQ

Member
Jul 11, 2020
5
0
well for the most part, you don't have to change the defaults. however, just like the BackgroundNeutralization tool, PCC needs to sample the background of the image to neutralize it properly. basically you can optionally define a preview over an area of the image that you know to represent the background only, and also set the threshold slider to tell PCC/BN how bright the brightest background pixels are. if your un-color-calibrated image has a huge disparity between brightness across the 3 channels, one of them will have really high signal in the background. if that level exceeds what is configured in the tolerance slider, PCC will ignore it, thinking it is signal and not background. so you should take a look at the background pixel values with the readout tool and set the slider just slightly greater than the level of the brightest background channel. if the field is really busy with stars and nebula it might be a good idea to set the background area of interest as well.
Thank you time and your reply. I followed your suggestion and under the PCC configuration so far i tried to configure a preview in the most representative area for the background (darkest one) and using statistics process I found the greatest level on brightest in each channel and configure in the PCC that value as you suggested. I did run the PCC again and get same results a strong green cast on my image. I'm running out of ideas , any opinion coming from you guys will be really appreciated. Thank you!

Rodrigo Quiroga
 

Attachments

pfile

PTeam Member
Nov 23, 2009
5,622
104
can you post an .xisf of the original image with no modification (that is, the xisf immediately out of WBPP or ImageIntegration.)
 

wrichards

Well-known member
Jun 22, 2020
69
0
DItto - exact same results here. I've attached a screen shot of the original image along with the original histogram, the image generated by the PCC process after linking the RGB channels and applying STF, the White Balance plots, and the histogram of the PCC image.
 

Attachments

wrichards

Well-known member
Jun 22, 2020
69
0
can you post an .xisf of the original image with no modification (that is, the xisf immediately out of WBPP or ImageIntegration.)
Absolutely - I've posted mine HERE. Thanks for your help! I'd really like to make sure I'm doing this step properly.
 

RodrigoQ

Member
Jul 11, 2020
5
0
can you post an .xisf of the original image with no modification (that is, the xisf immediately out of WBPP or ImageIntegration.)
Of course!!! Thank you so much for your support , I really appreciate your time and patience! As @wrichards I'm wanna make sure I'm doing it correctly.
Following the link to access to the XISF File: HERE

Thank you again!!!

Best Regards,

Rodrigo Q
 

pfile

PTeam Member
Nov 23, 2009
5,622
104
Absolutely - I've posted mine HERE. Thanks for your help! I'd really like to make sure I'm doing this step properly.
this one seems OK to me - i defined a very tiny preview on what i thought was probably background, but maybe higher in the right upper corner would be better. there is of course a lot of dust in this region which is probably not neutral in color so it is important to avoid that. i didn't process the image at all, that's just STF with the channels locked.

Screen Shot 2020-07-13 at 9.59.45 AM.jpg
 

wrichards

Well-known member
Jun 22, 2020
69
0
this one seems OK to me - i defined a very tiny preview on what i thought was probably background, but maybe higher in the right upper corner would be better. there is of course a lot of dust in this region which is probably not neutral in color so it is important to avoid that. i didn't process the image at all, that's just STF with the channels locked.

View attachment 8748
The Preview box I selected was a very dark section at the top-center, and the result was still very green. Was that a bad spot to select?
 

wrichards

Well-known member
Jun 22, 2020
69
0
this one seems OK to me - i defined a very tiny preview on what i thought was probably background, but maybe higher in the right upper corner would be better. there is of course a lot of dust in this region which is probably not neutral in color so it is important to avoid that. i didn't process the image at all, that's just STF with the channels locked.

View attachment 8748
I tried selecting an empty rectangle in the upper-right corner and it still turns the image green:
 

Attachments

pfile

PTeam Member
Nov 23, 2009
5,622
104
did you tune the background slider though? you need to do that for the values you see in the background area. if you use the preview bounds i used and the threshold i used, do you get the same result?

are you locking the channels before the STF? after PCC runs you need to recompute the STF...
 

Juan Conejero

PTeam Member
Sep 2, 2004
7,555
96
56
Valencia, Spain
pixinsight.com
These images can be calibrated without problems with our PhotometricColorCalibration tool. However, default parameters are suboptimal for these wide-field images because they lead to poor sampling of photometry data, so they must be changed to achieve good results.

Take a look at the following screenshots:

Lagoon/Trifid:

Rho Oph:

Default photometry parameters usually work well for most regions of the sky. However, wide field regions covering dense milky way areas pose particularly difficult problems, mainly because of the huge amount of stars and the complexity of the background.

As you can see, I have tweaked photometry parameters in order to ensure good sampling of the APASS catalog. For the Lagoon/Trifid image I have forced a limit magnitude of 12 and a photometric aperture of 8 pixels. With these parameters PCC has performed an absolutely perfect white balance calibration with about 3000 stars from the APASS catalog. The result is excellent, as expected from the white balance function graphs.

For the Rho Oph image I have used a limit magnitude of 10 and a photometric aperture of 6 pixels. These values are more appropriate for a wider field. The result is also very good, although APASS data for the r' filter shows much more dispersion. Despite that, the achieved white balance is excellent IMO.

As for background references, they are not very critical, as you can see in the examples above. Just try to select a relatively free area, such as a dark nebula, and avoid selecting background references too far from the center of the image if you have significant gradients. Use the Statistics tool to define a reasonable upper limit for background neutralization.
 
  • Like
Reactions: pfile

Juan Conejero

PTeam Member
Sep 2, 2004
7,555
96
56
Valencia, Spain
pixinsight.com
Rho Oph again with a more plausible background reference:
https://pixinsight.com/forum-images/20200713/PCC/Desktop5.jpg

This new result is much better, since the original background reference (Desktop4 above), having a red dominant, led to a blue cast. Anyway, background neutralization can be performed again after PCC. White balancing (multiplicative) and background neutralization (additive) are completely independent transformations.
 

wrichards

Well-known member
Jun 22, 2020
69
0
Juan,

Thanks again for all your helpful advice. I can now generate a reasonably color-calibrated image.

But this raises another question - if PCC is doing Background Neutralization, then is it even necessary (or perhaps harmful) to do DBE or ABE prior to doing PCC?
 

Juan Conejero

PTeam Member
Sep 2, 2004
7,555
96
56
Valencia, Spain
pixinsight.com
Fixing gradients before PCC is advisable, since gradients degrade the entire color calibration process. Of course, gradient removal must be done correctly or it may lead to worse results than not fixing gradients at all. As a general rule, never use ABE for images with scarce free sky areas like these ones. Use DBE with great care or, much better, use multiscale gradient correction when possible.

Background neutralization cannot remove gradients, since it is a single additive function (also known as a pedestal) applied to each channel of the image. A gradient is a surface function that must be modeled for each pixel of the image.
 

wrichards

Well-known member
Jun 22, 2020
69
0
Fixing gradients before PCC is advisable, since gradients degrade the entire color calibration process. Of course, gradient removal must be done correctly or it may lead to worse results than not fixing gradients at all. As a general rule, never use ABE for images with scarce free sky areas like these ones. Use DBE with great care or, much better, use multiscale gradient correction when possible.

Background neutralization cannot remove gradients, since it is a single additive function (also known as a pedestal) applied to each channel of the image. A gradient is a surface function that must be modeled for each pixel of the image.
Ah, OK. That explains what just happened. I applied DBE to my image (using what I thought were very carefully selected sample points), then applied the exact same PCC parameters to the DBE image and the result was horribly green.

I've never used Multiscale Gradient Correction - I guess I'll need to look into that process and learn it.
 

Attachments