Normalization problems with a big mosaic

@jkbsahner Hi Jakob. Can you tell us more about how you processed the 104 panels of the mosaic? Did you load all lights in one go or did you process them individually and then loaded the 104 integration results as lights to create the mosaic? I suppose the latter but please confirm.

Did you do any post-processing on the 104 panels before loading them to create the mosaic?

What settings did you use to create the mosaic exactly? Please describe them for all tabs 3 to 6. 

Would it be possible to share the 104 panels with us so we can have a look at them and help you solve this? We have a file server where you can upload the files so you don't have to upload them to Dropbox.

There may come more questions after we have received your answers 🙂

Thanks, Wouter

@wvreeven Hey Wouter,

I did no post-processing on the 104 images, just stacked all the data to get the 104 sigle images and cropped 7/104 slightly.

For the settings I used for integrating the low res version of the mosaic look at the images below

As I said I used the same settings on both version 1.083.0 and 1.083.4 but got a different (worse) result after integration on the 1.083.4 version...

I will upload the images on your file server

Greetings
Jakob

@jkbsahner Hi Jakob, is there any reason why you do not use "automatic" for "integrate" in tab 6? Have you tried it at all? If no, can you do that and see if that helps? Also, it looks like these are the settings for creating the mosaic. With what settings did you process the individual panels?

Well the integration has no influence on the overall colors of the image as far as my understanding goes. I dont choose automatic because each frame should have the same weight in the integration because its a mosaic and no normal stack. 

Back to the main problem: It seems like there are some differences in the version 1.083.0 and 1.083.4 because I get a different result as you hopefully saw in the images.

Is there a way to provide me a downloadlink to the original 1.083.0 version??

I think that would be the fastest solution for me. If there still is the same problem with the old version then I have to find a new fix for it 🙂

Interesting as there were no changes between 1.083 and 1.083.4 for specifically this (as far as I know). As for the integration setting, I would indeed try automatic first. Setting it to equal will cause a worse result if any of the frames have worse quality. Previous versions can be found here: https://www.astropixelprocessor.com/community/appreleases/downloads/

ps. 1.083.0 specifically isn't on that page indeed. If you need that one we need to ask Mabula.. it's a version which has a major bug in registration so not really advised.

@vincent-mod Well I would like to test it again with that version 🙂

I asked Mabula to answer here as well, hopefully this evening. 😉

@vincent-mod Thank you!

Posted by: @jkbsahner

I will upload the images on your file server

Did you ever do this? I guess not since I don't see them on our file server. Can you please upload the 104 images so we can help you get this solved for you? As Vincent indicated, there are no changes related to this between 1.083.0 and 1.083.4 so I doubt the changes in your integration results can be explained by the upgrade to 1.083.4.

@wvreeven No I didnt had the chance to do it yet. I will do it now!

Hi Jacob @jkbsahner @wvreeven and @vincent-mod,

To start with, if you are making a mosaic, the automatic integration settings will have equal weights because it is a mosaic! APP checks this and will apply different weights depening on the registration mode ;-). So the settings that you set : average with equal weights and no outlier rejection is exactly what APP will do in automatic integration of a mosaic.

Now to your issue with the colors. I have inspected the 3 images in your dropbox link and looked at all your settings.

The 1.083.4 does not look very different to me than the 1.083.0 version. How did 1.083.4 look with LNC enabled? I see no result of that.

We know that there is an issue with the following which can create an unwanted color shift:

1) mosaic registration

2) with advanced normalization

3) and background neutralization enabled

In this case, the normalization parameters are calculated only on parts where the images overlap. Now this can create an unwanted color shift if that area where normalization paratemeters are determined of the reference frame hold little sky background (meaning mostly covered by an object like a galaxy, nebula, milkyway part). Now it is very important that you realise that background neutralization is only a linear 3 factor operation which will never harm nor destruct your data. So in this case, I would think that simply running background calibration will solve your issue. Have you tried this? If not, can you please try that and let us know if that solves it?

Between the stable versions 1.083.0 and 1.083.4 for Windows, there has not been a single change that would affect this behaviour. So providing 1.083.0 for download will simply not solve your issue. And I am very hesitant to provide it, since it has a serious concurrency bug in the calibration engine.

Maybe you used a beta 1.083 version? between the beta's some changed were made in normalization having to do with background neutralization.

Mabula

Hey guys,

When you read this you can probably see the folder with all the images on your file server (all images should be uploaded by 2 pm)
Its called "Jakob Normalization Problem"

Im currently running a series of tests with different settings and the recommendations from @mabula-admin. I try to finish it all by tomorrow and send you the results.
I had a look on the images in my dropbox folder and yeah the are not massively different but I still prefer the colors of the first one. I'll try everything to get this look again...

I just want to have the best starting point after the full size integration. With my dataset and the settings I use it takes about 2,5 days to integrate so I dont want to waste that time!

Thanks again for the help!
Jakob

Well...

I just finished the first test and I dont know why but now it looks how I want it to look...
I dont know what or if I did something wrong the last time but now it looks good again 😀

The image you see below was done with the same settings I posted earlier. I still dont understand why it looked weird one time and 2 times it is good... Well I will let it run with LNC now and if the result looks good I will run it with full resolution!

Sorry, maybe it was just a misunderstanding or mistake on my part....

I will keep you updated!
Jakob