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.
I wish I could work out whats gone wrong here, but I ve tried a few different settings with the same result
it gets better 😫
I wish I could work out whats gone wrong here, but I ve tried a few different settings with the same result
Hi Tony @tonymacc,
Can you share your settings in 3) and 4) ? That might give me a clue on why it fails.
Kind regards,
Mabula
Hi Tony @tonymacc,
Thanks, in 3) analyse stars, can you try the following and report back if registration is then better? (settings in 4) register look good 😉 )
- disable automatic minimum stars target
- set kappa somewhere fixed on 3-10
- set automatic maximum star count lower, to 2000-5000 stars (10000 stars is really very high and will increase registration time and less stable results)
So the idea is that the star count is too high actually. Keeping the kappa value fixed will propably give you more consistent results per panel as well.
Aim to get a maximum star count of about 2000-5000 stars. I think that should give a stable or better result then.
Let me know what happens with these recommendations 😉
Mabula
@Mabula, almost, the top seems to be ok but now the others are not stitched, Kappa was at 3 and max stars was 3500. I will try another setting.
kappa 003 and 2000 max stars, an improvement
I tried kappa 003 and max stars 1700 with much the same as last post, I cannot get the kappa lower than 003, tried max stars at 1400 and it failed to register most of the panels.
this is the result of kappa 003 and max stars @ 1700, any other settings I can try @Mabula?
@Mabula, Whilst looking around the forum I think I might know why mine is not working. I found this from another post
I assumed mbb was only to be used when stitching, so looking at the post in the picture it looks like I should have used mbb in the individual stacks?
So I redone the stacks individually, 3)kappa 003 and max stars at 2000, disabled min stars. 4)quadrilaterals start 1 stop 5, use dynamic distortion correction, untick same camera and optics, registration mode normal and model set to projective. 5) advanced set in mode. 6) from the top...equal, average, full mode, Lnc 1st 3 iterations, mbb 10%, sigma clip kappa 3 with 1 iteration, lancos-3, no under-overshoot.
Then for the mosaic 3) kappa 003 max stars 2000, 4)quads start 1 stop 10, use dynamic distortion correction, untick same camera and optics, registration mode to mosaic and model is projective. 5) advanced mode. 6) equal, average, full, lnc 2nd, 3rd and 4th used to same result, mbb from 10% to 15%, sigma clip kappa 3 with 1 iteration, lancos-3, no under-overshoot.
I tried kappa 003 and max stars 1700 with much the same as last post, I cannot get the kappa lower than 003, tried max stars at 1400 and it failed to register most of the panels.
this is the result of kappa 003 and max stars @ 1700, any other settings I can try @Mabula?
Hi @tonymacc,
I think that looks very good 😉 I can't see all details on the screenshot, but from what I can see, registration looks good, doesn't it?
Mabula
@Mabula, Whilst looking around the forum I think I might know why mine is not working. I found this from another post
I assumed mbb was only to be used when stitching, so looking at the post in the picture it looks like I should have used mbb in the individual stacks?
Hi @tonymacc,
Yes, indeed, I would recommend to use MBB in almost all cases 😉 It simply reduces stack artefacts at the borders of your integrations as well, besides helping to blend the mosaic panels. The individual panels created with MBB will have the benefit, that you will usually won't need to crop the panels before the mosaic integration. If the mosaic panels have stack artefacts, that could upset the LNC function as well if you create a mosaic.
So use MBB and a low level LNC to greatly reduce stack artefacts at the borders of your mosaic panels and then create the mosaic with LNC and MBB 😉
Mabula
So I redone the stacks individually, 3)kappa 003 and max stars at 2000, disabled min stars. 4)quadrilaterals start 1 stop 5, use dynamic distortion correction, untick same camera and optics, registration mode normal and model set to projective. 5) advanced set in mode. 6) from the top...equal, average, full mode, Lnc 1st 3 iterations, mbb 10%, sigma clip kappa 3 with 1 iteration, lancos-3, no under-overshoot.
Then for the mosaic 3) kappa 003 max stars 2000, 4)quads start 1 stop 10, use dynamic distortion correction, untick same camera and optics, registration mode to mosaic and model is projective. 5) advanced mode. 6) equal, average, full, lnc 2nd, 3rd and 4th used to same result, mbb from 10% to 15%, sigma clip kappa 3 with 1 iteration, lancos-3, no under-overshoot.
Hi Tony @tonymacc,
The registration of the mosaic is fine now, isn't it?
What is the problem now you think? (the post is about the mosaic not stitching)
The illumination differences that we see can probably be fixed with the Remove Light Pollution (LP) tool, have you tried?
You can try to correct the individual panels for gradients/LP first before creating the mosaic.
Kind regards,
Mabula
Thankyou @Mabula. 😉 still a learning curve for me.
Thankyou @Mabula. 😉 still a learning curve for me.
Hi @tonymacc, excellent, the mosaic does look great already 😉 Is this your first of this size?
I really like this Region in H-alpha, so much interesting structures in the nebulosity !
Since there is so much nebulosity in your data, the current LNC implementation can sometimes give waves in the sky background, as you can clearly see in your latest result.
In my experience, the best way to deal with this with a lot of control, is to correct the individual panels first with the Remove Light Pollution tool before creating the mosaic. In that fasion, I created this 9 panel in H-alpha and RGB (so 2x9=18 panels) mosaics using data of Yves van den Broek:
This dataset actually was "the source dataset" for me to invent the LNC algorithms in APP. I will upgrade the LNC algorithms pretty soon though, because I know it isn't working perfectly all the times 😉
Mabula
Sort of yes, started this 2015 😆 I have all the oiii and am almost finished collecting the sii. I have already done this mosaic in ha in pi so I know it goes together nicely but it just takes so so long.
Great work, sticking to a project to be able to finish it at some point 😉
Do you mean the mosaic registration takes long, or perhaps the LNC calculation?
Mabula
@Mabula, Well I've given it quite a few goes over the last few weeks but still I cannot get a clean merge of some of the panels, I am at a loss now what to do next.
@Mabula, Well I've given it quite a few goes over the last few weeks but still I cannot get a clean merge of some of the panels, I am at a loss now what to do next.
@tonymacc, can you share the best result that you created sofar, so we can have a look at why it's not good yet?
Registration was no problem right? So I assume that you struggle with getting the sky background flat?
Mabula
@Mabula, Well I've given it quite a few goes over the last few weeks but still I cannot get a clean merge of some of the panels, I am at a loss now what to do next.
@tonymacc, can you share the best result that you created sofar, so we can have a look at why it's not good yet?
Registration was no problem right? So I assume that you struggle with getting the sky background flat?
Mabula
the file is too big for posting
@Mabula, Well I've given it quite a few goes over the last few weeks but still I cannot get a clean merge of some of the panels, I am at a loss now what to do next.
@tonymacc, can you share the best result that you created sofar, so we can have a look at why it's not good yet?
Registration was no problem right? So I assume that you struggle with getting the sky background flat?
Mabula
I can do this in another program but seem to fail here, stitched a 4 panel and 6 panel no problem but struggling with this. LNC 4th and 10 seems to give the best result using sigma clip, above and below that its not as good.
3/ unticked min stars, max set at 2000, kappa lowest it will go is 003.
4/ quadrilaterals, start 1 stop 10, use DDC, untick same camera, registration mode mosaic, model projective,
5/ advanced,
6/ lnc 4th and 10, sigma clip,
It still isn't clear what the problem is Tony. This looks rather fine to me.
In the screenshot, the registration looks good and I don't see mosaic seams.
There is plenty of room to correct the background i think with the Remove Light pollution tool.
Can you formulate what the problem is in your opinion?
Perhaps, you assume the LNC should give you a perfect restult out of the box? That will not happen. You still need to correct residual gradients.
LNC only makes sure that illumination is balanced in the overlapping regions, it doesn't make a perfect background (yet).
Off course, I can demonstrate how I would do this if you share the mosaic panels.
Mabula
Hi Tony @tonymacc,
I have downloaded them and will give it a try now 😉
Mabula
👍 @ Mabula, I am using version 1.059 I think, if that helps. 🙂