There is a part of me that thinks it might be in the individual frames because it shows up on the outside edge of the mosaic where there is no overlap between them. I could understand the overlapped areas looking slightly different but not the outside edge. If you can download the 21 stacks I would be willing to try the mosaic myself to see what happens. When I needed to give Mabula some photos he gave me this information to do that. I hope I am not breaking any rules here. Just go to this site and use the username and password to enter and make a folder with your name on it so we will know it is your data.

https://upload.astropixelprocessor.com/

username- uploadData

password- uploadTestData

Tom

First of all, thank you both for taking the time to look into my problem and for sharing your experiences and suggestions!

@John: Your suggestion to apply APP's Light Pollution Reduction (LPR) to the individual panels before mosaic registration sounds promising. I'll test this on four adjacent panels with particularly noticeable transitions before reprocessing the entire dataset.

Your point about Mabula advising against splitting large mosaics into smaller sections and then combining them is also very helpful. I'll therefore avoid pursuing that approach.

@Tom: I find your observation about the rounded dark areas in the starless version particularly interesting. These could indeed indicate residual vignetting or an issue with the flat-field calibration.

All panels were calibrated using flats. However, what puzzles me is that the individual integrated panels appear relatively uniform when viewed separately. The dark structures, along with the differences in brightness and colour, only become clearly visible when all 21 panels are combined into the final mosaic.

I'll therefore investigate whether these structures are already present in the individual linear stacks or whether they are introduced or amplified during normalization and mosaic integration.

I'd be happy to share individual integrated panels or even the complete dataset if anyone is interested in investigating the issue or experimenting with different processing approaches.

My next step will be to carefully examine four adjacent panels, verify the flat-field calibration, and then test LPR correction before mosaic registration.

Thanks again for your support! I'll keep you updated on the results of my further tests.

Best regards,
Jörg

When I look at the starless version, the rounded dark parts on the outside of each panel look like vignetting maybe. Any chance there could be something going on with the flat frame calibration? Could you show us a single panel or perhaps Mabula would let you download the 21 panels so we could experiment with them. Although I am not sure if my computer would handle a photo that large. I don't think I will ever try a mosaic of this size but I am curious to see if there is a solution.

Good luck,   Tom

I'm not really qualified to advise on a project of this scale, but I will say that when I experienced a similar issue with a smaller mosaic I got a good result by applying APP's light pollution reduction tool to the individual panels prior to mosaic registration.

My suggestion would be to pick the four panels with the worst transitions, do a careful LPR on each, and test if the issue goes away.

I would also add in response to your question 6 that Mabula has said that building up a mosaic using smaller mosaics (ie dividing a big mosaic into chunks) is not recommended.

good luck

JC

Hi everyone,

I’m currently working on a rather ambitious astrophotography project and would really appreciate some advice from experienced Astro Pixel Processor users and the developers.

The project

I captured a large 21-panel (7 × 3) mosaic of the Rho Ophiuchi region from Isabis, Namibia, under exceptional Bortle 1 skies.

The dataset consists of approximately 105 hours of total integration time, with around 5 hours per panel and 25% overlap between adjacent panels.

My imaging setup:

• Telescope: Askar SQA55
• Camera: ToupTek 24AC (full-frame OSC)
• Mosaic: 7 × 3 panels
• Overlap: approximately 25%
• Total integration: approximately 105 hours
• Location: Isabis, Namibia (Bortle 1)
• Processing: Astro Pixel Processor
• Computer: MacBook Pro M3 Max, 36 GB RAM

My processing workflow so far

I calibrated, registered and integrated each of the 21 panels individually in APP.

Afterwards, I loaded the 21 integrated panels into APP to create the final mosaic.

I’ve experimented extensively with different mosaic integration and normalization settings, including:

• Multi-Band Blending (MBB), including 25%
• Local Normalization Correction (LNC), including second-degree LNC with 5 iterations
• Different normalization approaches
• Different reference panels, including a central panel
• Full Image normalization
• Background neutralization and gradient correction
• Different mosaic integration settings

I’ve also been experimenting with Drizzle during the individual panel integrations. Due to APP’s practical image-size and memory limitations, I cannot currently process the entire mosaic at its maximum potential resolution.

The problem

Despite numerous attempts, I’m still struggling with visible transitions between the individual panels.

The attached image illustrates the problem.

There are noticeable horizontal and vertical structures along the panel boundaries, often appearing as yellowish lines or areas with different background brightness and color.

Interestingly, some of the overlapping regions appear to have a more uniform or cleaner background than the central areas of the individual panels.

This makes me wonder whether the normalization and gradient correction are treating the overlap regions differently from the rest of each panel.

Increasing the strength of LNC and other corrections can improve the visible transitions, but unfortunately it also seems to suppress faint dust and nebulosity.

And that is exactly what I want to avoid.

The Rho Ophiuchi region contains extremely faint and complex dust structures, and with 105 hours of exposure under Bortle 1 skies, preserving those structures is particularly important to me.

I’ve successfully processed smaller mosaics, including 4 × 4 panels, without encountering these problems to the same extent. The difficulties seem to become much more pronounced with this larger 21-panel project.

My questions

1. What would be the recommended APP workflow for a 21-panel mosaic with significant faint dust structures and 25% overlap?
2. Would you recommend applying gradient correction individually to all 21 integrated panels before creating the mosaic, for example using APP, GraXpert or PixInsight DBE?
3. Could normalization based primarily on overlapping regions explain why those regions sometimes look better than the central areas of the panels?
4. What combination of LNC, MBB and normalization settings would best preserve faint nebulosity while eliminating visible panel boundaries?
5. Could Drizzle integration followed by downsampling or resampling contribute to these artifacts, or is this most likely an independent normalization/gradient issue?
6. Is there a recommended strategy for processing such a large mosaic at full resolution, potentially by dividing the mosaic into smaller sections before the final integration?

My goal

I want to achieve a seamless mosaic with a consistent background, natural colors and, most importantly, without sacrificing the extremely faint dust structures captured during those 105 hours.

I’m more than willing to reprocess all 21 panels if necessary, but before investing another significant amount of processing time, I would really appreciate some guidance on the most effective workflow.

I’ve attached an image showing the current result and the visible transitions between the panels.

Any suggestions, recommended APP settings or insights from users who have successfully processed similarly large mosaics would be greatly appreciated.

Thank you very much for your help!

Best Joerg

Original image:

Stars removed

For comparison, here's an earlier image processed with APP version 2.0.0-beta44:

Thank you very much for your post.

The general levels stretch with default settings is not stretching anything, so that is the expected behaviour as is explained in the YouTube clip that Meredith @wongataa shared (thanks !).

You can use the general levels stretch to further process already stretched images, like still adding contrast or saturation, detail enhancement etcetera. The actual stretching in the general levels stretch uses the Black and White values and gamma (this is called levels in applications like Photoshop). Stretching by adjusting the black/white/gamma sliders can be useful still for non-linear data or for investigating/analysing flats for instance.

Processing/stretching of linear data is advised to be done with the new Adaptive Deep Sky Stretch or ADSS and gives you the most powerful and data adaptive options with simple sliders. You can also enter slider values manually. I am working to further enhance ADSS. An important slider to use and test is the color preservation slider. Depending on the data, it can give you much better color in the lowlights without needing to add additional color saturation.

Please let me know if you have additional questions.

Mabula

Just to add to all the advice from Tom and John (thanks guys!). 

Use one of the preset formula that are programmed into APP's RGB Combine Tool, like the SHO, HSO formula to make a composite of SII, H-alpha and OIII. 

Once you have found the formula that you like the most, simply tweak the parameters with the sliders step by step to get the most pleasing colors, which always depends a bit on the data quality and the actual object that you are processing.

Please share your result with us once you have finished processing 😉 or if you need more help.

Mabula

Wow, that is awesome and quite fast 😊  ! Did the final mosaic look good? 

Mabula

Thank you very much for the feedback, happy to hear that the new changes are working well for you ;-)