Mosaic color cast m...
 
Share:
Notifications
Clear all

[Solved] Mosaic color cast mystery


(@annehouw)
Main Sequence Star Customer
Joined: 5 years ago
Posts: 55
Topic starter  

Hi,

Normally I find it smooth sailing with APP. This time however, I cannot get my fingers behind a problem I have with a mosaic and I would like to get some suggestions

Below four panes of M45. Integgrated and light pollution corrected. I put them side by side in Photoshop for easier comparison to the end result. Saved from APP with dpp 15% BG, 5 sigma, 0% base. Looks OK to me.

Also attached, the mosaic integration result, straight out of APP. Same view settings. Suddenly, there is a huge color cast in the result, which I cannot explain.

The settings were all the normal mosaic ones (including 2nd order with 5 and even 10 interations) . I also tried quality vs equal weights in intergration, but that did not help. 

Suggestions welcome!

The four panes (individually pre-preprocessed in APP at an earlier date and and hand-positioned in Photoshop for easier recognition of M45 )

M45 images

The mosaic result

M45 Mozaiek test mbb25 equal St

Anne

This topic was modified 4 months ago 2 times by Annehouw

ReplyQuote
(@vincent-mod)
Quasar Admin
Joined: 5 years ago
Posts: 4727
 

I think this is because there still are differences in between the panels in the background, which is probably statistically still causing this pattern when making the mosaic. You should be able to use the Light Pollution tool to correct this nicely.


ReplyQuote
(@annehouw)
Main Sequence Star Customer
Joined: 5 years ago
Posts: 55
Topic starter  
Posted by: @vincent-mod

I think this is because there still are differences in between the panels in the background, which is probably statistically still causing this pattern when making the mosaic. You should be able to use the Light Pollution tool to correct this nicely.

Thanks Vincent,

I did a careful 2n round of LP removal on the individual frames. (One of the things that I noticed earlier is that using Background Neutralisation in the viewer on these individual frames makes the nebulosity go white and the background yellow, like in the modaic result...).

The mosaic output is not good (30% BG for extra clarity of the problem).

 

M45 2xlpc St

As an experiment, I disabled Background Normalisation in the normalization step. This gave another result, but still off.

M45 2xlpc noBN St

In all results, the pane on the bottom left looks the most normal. It is the pane with the least nebolosity.

Looking at the last result, it looks like 2nd order LNC is introducing strange gradients (?)

I did experiment with using 1st order LNC and that helped somewhat:

M45 2xlpc 1stdegrLNC St

 

Of course, I can LPC the last result aggressively and that gives a picture, but at a loss of dim details. I still got the feeling that APP is going wonky on this dataset 😶. Care to have a go at it yourself?

M45 2xlpc nonorm 1stdegrLNC 1 St globallpc

 

This post was modified 4 months ago 5 times by Annehouw

ReplyQuote
(@annehouw)
Main Sequence Star Customer
Joined: 5 years ago
Posts: 55
Topic starter  

....and finally, I did a mosaic integration without any LNC. The individual panes should now be left alone and only stitched. But they are not. ! pane is left untouched, the other three (with most of the blue nebulosity) have the strange color cast.

M45 2xlpc noLNC St

 

So, something is off in the pipeline.....


ReplyQuote
(@annehouw)
Main Sequence Star Customer
Joined: 5 years ago
Posts: 55
Topic starter  

....and so I went back into the pipeline and saved the normalized images...and there is the root of the problem. Here screenshots of two panes:

The first one looks normal (30% BG)

normalisation  1

The second one is way off

normalisation  2

Same for the third and fourth frame.

normalisation  3
normalisation  4

 

I tried disabling normalisation in the integration, but that gave a funky end result, which I will not show here.

 

This is as much of a diagnosis as I can do with my knowledge of APP. 

So, as a recap: The input was a set of individual frames that I had integrated and LP corrected earlier, with fine result. In the process of making a mosaic, three of the 4 frames got color shifted in the normalisation step. As a result, the final mosaic ended up with severe color irregularities.

Would appreciated further analysis. 

 

 

This post was modified 4 months ago 2 times by Annehouw

ReplyQuote
(@vincent-mod)
Quasar Admin
Joined: 5 years ago
Posts: 4727
 

That might also be an indication of the calibration frames having an issue maybe? What happens when you leave those out? And what version of APP are you using, just to be complete. 🙂 When normalization goes white, that's usually also an issue in calibration.

Unfortunately I'm not able to process data for about a week, so maybe @mabula-admin or @wvreeven could have ago...


ReplyQuote
(@annehouw)
Main Sequence Star Customer
Joined: 5 years ago
Posts: 55
Topic starter  

@vincent-mod

I did not use any calibration frames in this mosaic processing as these images had been integrated as individual images (and corrected for LP) before. The result of that is in the very first (composited for convenience to see the panes, the overlap and the orientation) image at the top of this thread: Normally colored images.

APP version is the latest and the greatest. 1.083 😉 

 

There is no specific hurry. I need to acquire more hours on this subject. But this is a strange effect that I have never encountered before, and do not understand. For that and for future reference, I would like one of you to take a look at this with a better trained eye. 

This post was modified 4 months ago 2 times by Annehouw

ReplyQuote
(@vincent-mod)
Quasar Admin
Joined: 5 years ago
Posts: 4727
 

Love to have a look as well, maybe a good idea to already upload some of the original subs (like 10 for each panel) and your master calibration frames AND your 4 integrated panels on which you see this behavior.

Go to https://upload.astropixelprocessor.com and for the username and password, use: upload

Create a directory named “annehouw-mosaicIssue” and upload in there. Thank you!


ReplyQuote
(@annehouw)
Main Sequence Star Customer
Joined: 5 years ago
Posts: 55
Topic starter  

Hi Vincent,

Would like to do that, but the link you provided is not working:

Sorry, de pagina die u zoekt kan niet weergegeven worden


ReplyQuote
(@wvreeven)
Galaxy Admin
Joined: 4 years ago
Posts: 1647
 

@annehouw Please try again. Occasionally loading the URL fails if the server is too busy. Sorry for the inconvenience.


ReplyQuote
(@annehouw)
Main Sequence Star Customer
Joined: 5 years ago
Posts: 55
Topic starter  

@wvreeven

Thanks. Uploading at 1.1 MB/s as we speak. 


ReplyQuote
(@annehouw)
Main Sequence Star Customer
Joined: 5 years ago
Posts: 55
Topic starter  

Done!

Thanks for looking into this.


ReplyQuote
(@mabula-admin)
Quasar Admin
Joined: 5 years ago
Posts: 3117
 

Hi Anne @annehouw,

The upload server was undergoing some work and updates in the past couple of days. All finished now 😉 so it should be online with high reliability again.

Thank you for sharing this odd behaviour. I will definitely look at the issue and will report back with my findings.

I know that LNC can become unstable with high dynamic range objects on the mosaic borders like you are having here. But normalization failing here without LNC and even using advanced normalization definitely should not happen.

Mabula


ReplyQuote
(@annehouw)
Main Sequence Star Customer
Joined: 5 years ago
Posts: 55
Topic starter  

Hi @Mabula-Admin

I did some further investigation with interesting results, which I would like to share here.

I reprocessed all subs from the ground state. After that I did a careful light pollution removal. Because of the large areas of nebulosity, I can only correct for a part of the image. But this produced 4 nice individual frames.

 

On to mosaicing. This is with "neutralize background" disabled in the normalization step:

M45 2nd LNC

Colorwise, no wonky stuff, with gradients. This is with 2nd order LNC. (5 or 10 iterations..no big difference)

Now I went from 2nd order LNC to 1st order (3 iterations)

M45 LNC 1st 3it

Smoother results

Even went a bit further with 1st order LNC and 1 iteration. This was a way too much because the light level differences in the panels were showing a bit.

 

To anwer why  I got those strange colors in the first place,  I did a Mosaic with my "standard"settings of "Neutralize background" in the normalization step.

M45 Neutralize background

Looks familiar. 

For some reason, background neutralization has a difficult time with these large areas of bright nebulosity.

 

Going back to my earlier result, using the Remove Light Pollution Tool gave me a nice result for the integration time (1 hour per panel), shown at 20% BG:

 

M45 LP

 

I will document the findings in this  in my own notes as it is something that is not obvious from the start:

  • Disable Neutralize Background in normalize
  • Go easy on LNC and experiment (this is in the tooltips btw)

 

Anyway, I hope this is interesting for other users.

 

 

 

 

This post was modified 4 months ago 3 times by Annehouw

ReplyQuote
(@mabula-admin)
Quasar Admin
Joined: 5 years ago
Posts: 3117
 
Posted by: @annehouw

Hi @Mabula-Admin

I did some further investigation with interesting results, which I would like to share here.

I reprocessed all subs from the ground state. After that I did a careful light pollution removal. Because of the large areas of nebulosity, I can only correct for a part of the image. But this produced 4 nice individual frames.

 

On to mosaicing. This is with "neutralize background" disabled in the normalization step:

M45 2nd LNC

Colorwise, no wonky stuff, with gradients. This is with 2nd order LNC. (5 or 10 iterations..no big difference)

Now I went from 2nd order LNC to 1st order (3 iterations)

M45 LNC 1st 3it

Smoother results

Even went a bit further with 1st order LNC and 1 iteration. This was a way too much because the light level differences in the panels were showing a bit.

 

To anwer why  I got those strange colors in the first place,  I did a Mosaic with my "standard"settings of "Neutralize background" in the normalization step.

M45 Neutralize background

Looks familiar. 

For some reason, background neutralization has a difficult time with these large areas of bright nebulosity.

 

Going back to my earlier result, using the Remove Light Pollution Tool gave me a nice result for the integration time (1 hour per panel), shown at 20% BG:

 

M45 LP

 

I will document the findings in this  in my own notes as it is something that is not obvious from the start:

  • Disable Neutralize Background in normalize
  • Go easy on LNC and experiment (this is in the tooltips btw)

 

Anyway, I hope this is interesting for other users.

 

Hi Anne @annehouw, I was just about to have a very good look at your data, but now I fully understand what was happening with Background Neutralization on on your data.

It is not an error by any means then. The Background Neutralization option never harms your data, it is only there to give color data a clearer view of what is in the data. Sometimes the color data sky background can be very colorful depending on sky conditions and Background Neutralization can then help to better see what is in the data 😉

The Background Neutralization operation is very naive in the sense that it only looks at the whole histogram of the data and it works linear and multiplicative. So with your Bright Blue Nebula covering a large part of the Field Of View, the remainder of the sky turns yellow, opposite of blue to make it neutral over the entire Field Of View.

Since you used advanced normalization only normalization parameters are calculated on the mosaic borders in your case where the bright Blue Nebula is.. that then causes the real sky background to turn yellow since it had only normalization parameters with blue nebulosity in there...

I have made a note of this issue on my ToDo List to see if we can improve this behaviour. We need to make it a bit smarter for a case like this obviously.

Thank you very much for your clear explanation and your own trouble shooting.

Mabula


ReplyQuote
Share: