15th Feb 2024: Astro Pixel Processor 2.0.0-beta29 released - macOS native File Chooser, macOS CMD-Q fixed, read-only Fits on network fixed and other bug fixes
7th December 2023: added payment option Alipay to purchase Astro Pixel Processor from China, Hong Kong, Macau, Taiwan, Korea, Japan and other countries where Alipay is used.
Hi all,
There clearly is some confusion in the community about what a mosaic technically is and what not I think, so this FAQ is created to give a better understanding of what a mosaic is and when to approach it like that in registration.
Several users, especially users that have used Deep Sky Stacker (DSS) a lot, (but also other applications) are using APP's mosiac registration mode when it's not needed at all 😉 It will only complicate things a lot and it will slow things down hugely if you need to register and integrate 100s of frames.
DSS' mosaic function is really not a mosaic as far as I am concerned.
It's a composition mode that just shows all pixels in the integration of all frames. But DSS makes no attempt whatsover to register a mosaic. It simply can't.
The DSS mosaic setting has no influence on the registration so calling it a mosaic is bad naming (I think).
In APP this is called the full composition mode in 6) integration, which is used by default.
So what's a mosaic then?
Technically, a mosaic is a dataset that has some frames that have 0% overlap with the chosen reference frame. Then you can call it a mosaic technically and the registration can only solve that in the mosiac registration mode. A dataset like that can not be registered by DSS at all.
So if you have data that is spread over a large field of view, but where all frames have some overlap with 1 of the frames in the dataset (so that would be one of the frames in the center of the field of view), then this can and should simply be registered in the normal registration mode without using the mosaic mode.
Here is good example, this would be called a mosaic by a lot of DSS users, but it really isn't a mosaic and it was registered in APP using the normal registration mode.
It's data shot with a Nikon D5100 and a 50mm objective using a simple tripod without tracking. Iso 1600 and exposure of only 2.5 seconds.
So it's a wide field of view, a nice panorama, but
all frames have overlap with the frames in the center of the field of view, so this is technically NOT a mosaic ! and approaching it as a mosaic will only complicate things without a reason to do so.
in 4) register, use,
- same camera and optics on
- dynamic distortion correction on
- registration mode normal (so not mosaic!)
- projective model
Registration is then fast, optical distortion correction is very accurate as well, even with 100s of frames.
The first image shows registration & integration of 20 frames of the dataset, the other 2 images show 200 frames.
THIS IS NOT A MOSAIC 😉
So what's a mosaic then ?
A mosaic has frames that don't have overlap with whatever frame you choose as a reference in registration:
This then clearly is a mosaic. several mosaic panels have 0% overlap with whatever frame you choose as a reference:
Data courtesy of André van der Hoeven @andre_van_der_hoeven , H-alpha exposure in Cygnus:
THIS IS A MOSAIC 😉
Addtional and very important : if you do have a mosaic, don't try to register all frames of all panels all at once, that is very inefficient and not as robust as it could be!
If you have 9 panels of 30 frames each = 9x30 = 270 frames,
then registering 270 frames in the mosaic registration mode will take very long ! as some users have experienced. This is not efficient and also not the best method for precision.
You need to register (normal registration mode) and integrate the 9 panels separately first using both LNC and MBB in integration. LNC and MBB will strongly prevent any stack artefacts so chances are good that you won't need to crop (more data is always better for robustness of registration and SNR over the whole mosaic) any mosaic panel for a good mosaic registration & integration 😉
The integration of the mosaic itself needs to be done with LNC and MBB as well to prevent seams and illumination differences in the mosaic.
My recommendations currently would be:
For the individual mosaic panels:
- LNC 1st degree 3 iterations
- MBB at 5-10%
For the mosaic:
- LNC 2nd - 4th degree 3 iterations (only increase iterations and degree if needed) A higher degree can cause more illumination waves in the mosaic. (LNC upgrade for this is being worked on)
- MBB at 10-25% (depends on the amount of overlap between the mosaic panels, 10% overlap -> 10% MBB, 20% overlap -> 20%MBB)
Then the mosaic calculation is reduced to only 9 frames and calculating that is done in a couple of minutes on most hardware. So that will reduce the amount of time needed for registration greatly and it will even give better and more robust registration for your mosaic 😉
Kind regards,
Mabula
So first make a stack per panel using LNC and MBB. Then make the Mosaic with no LNC/MBB? With LNC/MBB? I am going to start over with the processing of my 72 Panel Orion Mosaic and need to know the best method before making the 72 stacks....
So first make a stack per panel using LNC and MBB. Then make the Mosaic with no LNC/MBB? With LNC/MBB? I am going to start over with the processing of my 72 Panel Orion Mosaic and need to know the best method before making the 72 stacks....
Hi Kees @kees_scherer,
Yes, integrating the individual panels with both LNC and MBB is highly recommended. If you don't you will need to crop the created panels to remove integration artefacts and the more data that you have for the mosaic, the better 😉
For the individual panels, use LNC 1st or 2nd, but not higher. The current LNC implemtation might gave waves with a higher degree, which you don't want, especially in the individual panels.
I have added the following information in the main post to be more complete:
The integration of the mosaic itself needs to be done with LNC and MBB as well to prevent seams and illumination differences in the mosaic.
My recommendations currently would be:
For the individual mosaic panels:
- LNC 1st degree 3 iterations
- MBB at 5-10%
For the mosaic:
- LNC 2nd - 4th degree 3 iterations (only increase iterations and degree if needed) A higher degree can cause more illumination waves in the mosaic. (LNC upgrade for this is being worked on)
- MBB at 10-25% (depends on the amount of overlap between the mosaic panels, 10% overlap -> 10% MBB, 20% overlap -> 20%MBB)
In addition to the LNC implementation, as you know, a new LNC implementation will come using several adjustments that should prevent the waves from appearing in the higher degrees. And I will work on making the calculations faster.
Kind regards,
Mabula
Stack of 11 sets of 300 frames, cropped for processing.
Each frame is 1.0 seconds exposure.
First pass registration using best frame from centre i.e. somewhere around 150 mark for that set.
14450 x 1650
Splodgy bit is the Rosette Nebula.
Used MBB at 5% for first integrations, 25% for mosaic. Still some artefacts.
Hi,
I'm working on a 2 panel mosaic and have come across this thread. Most of it makes sense but I would like to check a couple of things please:
panel 1 & 2 images each taken on 5 separate nights (so 5x multiple panel 1 and 5x multiple panel 2 images). Would the stacking procedure be as follows:
Stack night 1 panel 1 with master calibration files and LNC 1st degree 3 iterations, MBB 5-10%
Stack night 1 panel 2 with master calibration files and LNC 1st degree 3 iterations, MBB 5-10%
Same for nights 2, 3, 4 and 5
Then stack all panel 1's together for a panel 1 master, is this also LNC 1st degree 3 iterations, MBB 5-10%? and do I need to include calibration files again?
stack all panel 2's together, assume same settings as for panel one master
Finally stack panel 1 & 2 masters with LNC 2nd-4th 3 iterations and MBB 10% and mosaic registration mode set
Thanks
Ed
You can do it like that indeed, if you have flats for each night that is a nice workflow. Otherwise you may combine all data for panel 1 and use the same calibration files. When each panel is created and corrected, that is all there is to it. You can load each panel again as lights (first clear the data in APP), then setup APP for a mosaic and create the mosaic (so without calibration files). Mosaic settings;
- Amount of stars: 2000 or more
- Registration mode: mosaic
- Normalization: Advanced
- LNC 3, it 1 (to begin with)
- MBB: amount of overlap (if you have 10% overlap: 10%)
Thank you Mabula for the continued improvement of APP.
I have a 6 panel mosaic which combines 270+ frames that I need to process. In my haste (and before finding this thread) have tried to register all frames together and the registration phase went on for over 24 hrs+ and then failed. 😪
I use a OSC with a quadband filter, Ha OIII & SII and stack using algorithms Ha OIII extract Ha, HA OIII extract OIII & Ha OIII mono then RGB combine these results in tools, remove light pollution and alter colours using HSL selective colour.
So, have I got this right in that I need to do this routine for each of the 6 panels, creating 6 new combined and adjusted files and then use these to create the mosaic? Or, do I stop short after removing LP, then after combing them into a mosaic then adjust the result in HSL selective colour.
So to then create the final mosaic image with these 6 new files, do I then put those images through the same tabs 3, 4, 5 & 6? If so, would that not cause registration issues?
Thank you, just trying to get my workflow clear in my head.
Julian
So there are multiple ways you can do this, but I'd suggest to indeed make the panels first, even combined with all the data and then create a mosaic. The integrated files will be registered to each other (with the right settings) and the mosaic created.
Thank you.
Had a go with a quick process and it worked just fine , thanks again
Just waiting for noise reduction tool and we are good to go.
I still struggle to understand combine RGB with SHO/HOO for my colour CMOS and the slider settings plus Selective colour adjustments and which channel to adjust to get decent colours to pop out?? A tutorial would be fab if there are any or one to follow perhaps?
Keep up the fab work
Are there any settings that might allow efficient merging of a mosaic from a Vaonis Vespera, where the mosaic is made by going around in something like a spiral?
The attached video shows an animation of the subs from a mosaic sequence. Typically there'd be quite a lot of overlap between frames that are relatively nearby in the sequence, but some frames may have no overlap at all with any one reference frame.
With a test image that had 578x 2MP subs this required over 160,000 registration pair attempts, and a 137 GB working file!
The Vespera does already stack these onboard very handily itself with something they call "CovalENS" mode, but I want an off-board method to stack with more control over which subs are used for those (rare) cases when it fails. For example, I've sometimes had planes make it through into the generated stack.
To APP it won't be any different than a regular mosaic, no matter what way it was made. The same algorithm will be used, the latest beta seems to be faster and more memory friendly at least. To avoid trails, each panel needs to be made separately, so if the trails end up in a panel, that panel needs to be addressed either by clipping or removing the image with the trail in it.
