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.
Lets' talk again when you found time to look it thru, yes!
APP is really great with mosaics and a few improvements in the way of PTGUI would make it a tool noone can ignore I believe. people still do a lot with wider angle lenses, yet how many hours would you need to do this below with a let's say a Sigma Art 50 which some want to stop down to f/4 due to coma?:
this is a 10 hour mosaic done with APP using 8 panels of a 135mm Sigma Art lens at f/2.2. I applied starnet++ in Pixinsight afterwards and enhanced structures further with PS, Nik colection plugins etc.
Doing mosaics with lenses like the Sigma Art 105 or 135 or the new Sony 135mm are really powerfull and the way to go I think...
Well he got the message, but I have no control over his time unfortunately. I sincerely apologize. Things are very busy at the moment with the imminent release of 1.076 and Astrofest in London, I'll make sure to bring it up again when we're back.
I did talk with Mabula about it at Astrofest, but it seems the problem lies in the data itself. For some reason the data you start with seems to have an issue which makes it hard to correct well. I'll have a look at it myself as well, I hoped Mabula was able to do so in between, but he seems too busy with the Beta and stable release at the moment.
Hi Vincent, I understand that you have a lot going on... actually it has been like this since 2017 most of the time...
I do not really see why the Nikon D750 should be the cause, since the single frames look good and frames done under same conditions show different color. What makes you believe it is the data from the D750?
The reason I suspect that, is that APP usually is doing a very good job in correcting gradients and color problems. When things don't want to work out it usually (just by experience in helping here) is something in the data. Ofcourse, this is a wide mosaic and that's not made that often so there could be something going on still in APP, but it would be the first time I see that to this extent. Also I saw a (beautiful) mosaic you created at Astrofest, the starting panels also had a lot of color issues. This, APP was able to correct very nicely and the result was beautiful, but that also made me wonder if in this case, the data is causing an issue. I will however try it myself as well.
The reason I suspect that, is that APP usually is doing a very good job in correcting gradients and color problems. When things don't want to work out it usually (just by experience in helping here) is something in the data. Ofcourse, this is a wide mosaic and that's not made that often so there could be something going on still in APP, but it would be the first time I see that to this extent. Also I saw a (beautiful) mosaic you created at Astrofest, the starting panels also had a lot of color issues. This, APP was able to correct very nicely and the result was beautiful, but that also made me wonder if in this case, the data is causing an issue. I will however try it myself as well.
I see. You believe the problem is not APP so less priority? To me it seems APP is causing the problem since in the single panels there is no gradient visible for me, see above. I am not sure which mosaic you refer to. At Astrofest? Old data? Except of this milkyway mosaic I think I did not show any mosaic and single panels recently.
Sorry if I gave that impression, ofcourse I try to treat everyone exactly the same regarding priority. In this case it's a bit more complicated however and the only thing I try to do is to figure out where to look for the cause. And I take into account all possibilities, like I said it could be APP, but it could also be the data. Not seeing any gradient by eye, doesn't necessarily mean the panels are the same (taking everything into account). But, I'm investigating as we speak, starting with a portion of the panels first.
So I started looking at the stacked panels, first thing I always do is to go in and correct for light pollution. For this I over-saturate and over-stretch the panel to really bring out the gradients in the background. So this image below looks worse than non-stretched, but it's obvious there are multiple gradients in the background. Where they are coming from I don't know (partially light pollution on the lower left towards the horizon), but there seems to be more going on. Maybe a bit of reflection? But this background is very challenging to get corrected. I'm going to give it my best shot, but I think this is pointing towards the issue, having an uneven background, which might also change when you take pictures of various parts of the sky, will pose a challenge to really get right across all panels.
But, giving it a go and if I get it to work I'll post my workflow.
Second panel I looked at shows a very similar pattern, it definitely looks like there is a light pollution problem on the lower left, causing internal reflections in the lens. As you can see this weird band of light on the lower left, followed by a darker band, then another band of light. Or very maybe faint clouds that reflect a bit of sky background, do you remember if that could've been the case?
So I started looking at the stacked panels, first thing I always do is to go in and correct for light pollution. For this I over-saturate and over-stretch the panel to really bring out the gradients in the background. So this image below looks worse than non-stretched, but it's obvious there are multiple gradients in the background. Where they are coming from I don't know (partially light pollution on the lower left towards the horizon), but there seems to be more going on. Maybe a bit of reflection? But this background is very challenging to get corrected. I'm going to give it my best shot, but I think this is pointing towards the issue, having an uneven background, which might also change when you take pictures of various parts of the sky, will pose a challenge to really get right across all panels.
But, giving it a go and if I get it to work I'll post my workflow.
Second panel I looked at shows a very similar pattern, it definitely looks like there is a light pollution problem on the lower left, causing internal reflections in the lens. As you can see this weird band of light on the lower left, followed by a darker band, then another band of light. Or very maybe faint clouds that reflect a bit of sky background, do you remember if that could've been the case?
hi Vincent,
the data was taken in Namibia and some pics close to the horizon show (of course) some lightpollution. my daughter wanted to take the entire MW, in Germany we are done with the rest now and would like to finish a 360 degree picture... not so easy it seems.
I did a mosaic (with LP and starcolor corrected panels done in APP, exported with sat 35 as tiff) with PTGUI pro:
For the first try it looks pretty good, even though I can still see panels with colors that differ from the rest. but this seems normal. APP is rendering:
Done with unprocessed panels it looks worse:
I uploaded the LP/color corrected panels into my cloud, now... maybe try with these:
Thanks for uploading the other panels Stefan. Would definitely be nice to get this to work, I think it really shows the crazy dark skies overthere. I'll download them later, first trying to get this previous dataset to behave. What kind of projection did you do in APP btw?
I have no experience with PTGUI, so no idea what that really does. APP is actually going for getting the entire mosaic to have as equal background as possible. I think it might be the case that some of the gradients are throwing it off a bit. I'll discuss my findings with Mabula as well.
since last summer I had serveral attempts with this. I guess I tried everything. From my protocol I see the following:
stars >800, triangles (or pentagons, lasted forever if not right, sometimes qualityscore really off for no reason, so counting stars crucial sometimes, trial and error, but it worked well in the end...), scale 5- 15, mosaic defaults except calibrated projection (40mm, 5,95 mju), equit. proj. reference set so that it is horizontal. downscaled to 0.1 for faster results
Thanks for that, I'm continueing the quest. 😉 My own computer is getting a bit old for these huge mosaics, so I'll also have to downscale. I've corrected your frames as well as I could, now trying to produce something similar to your mosaic.
Ok, got to the same point as you. Individually I can correct them realitively well, but they do still contain gradients. When creating the mosaic, a whole new background-model is made because the field of view is now across all panels, and because the panels are different in quality, I think this is causing an issue in that sense. However, I'm going to keep working on it to see if this can become better.
ps. I do find the creation of the mosaic itself pretty awesome, this one wasn't possible some versions ago. 😉
And this is about the best I can get it myself. The problem with scaling down in this case is that the resolution becomes so low, APP is not able to detect stars properly at 0.1. I processed this on scale 0.5, maximum stretch and saturation. First tried to find the light polluted parts of each panel and trying to correct for that a bit by placing boxes at those places. I did this twice, saving in between. Then I did a color calibration in which I moved the green to magenta as there is a clear green cast in the data. After that I turned my extreme stretch down and went into the light pollution correction yet again, turning the stretch down now showed me a more general gradient, so I selected a few big boxes across the different parts of what I saw and it corrected it to the image I uploaded here.
The biggest issue in the data is the fact that the data is far better in the middle and you loose a lot of contrast on the right hand side, on the left there is more light pollution from the horizon. These differences seem to interfere with a very nice correction.
But I do think, maybe with some other software like PS (and maybe in the future in APP, who knows), you can now see these 3 distinct parts, select them and play around a bit with contrast and such on those parts.
Thanx Vincent. All in all it looks like my serveral attempts AND that even how hard you try it will not really work satisfactory, right? Yes, the gradients close to the horizon ... But they are a lott better processed in the single panels. Any remarks on this from Mabula?
One other thing concerning big mosaics: does APP use swapfiles? The need to downscale these mosaics is not really neccesary on some issues. A file the size of like 8gb is hardly possible to handle. with pixinsight no problem.
Regarding removing the gradients I think this one is hard yes. It will be better in a future version of LNC which Mabula wants to implement, but as it's not here yet that doesn't help you right now. If you have PI already, you might check if there is a tool to correct the mosaic background a bit better in there? Then you will have the full mosaic work done in APP at least which will be the best regarding the creation of one and maybe some tools from PI for post processing.
APP doesn't use swap files no, better or different memory management might be looked into, but at the moment it does its own based on actual memory as the penalty in speed for instance would be rather big. These are things we're looking into constantly as well though, next to speed.
