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,
I am having a lot of issues getting AstroPixelProcessor to process a mosaic. I've got a six panel mosaic of EtaCarina done with a one shot colour camera. I've been using APP for about 10mths and can process individual images at a basic level now, so I've some of the basic's down. I've got 6 x 25subs, with a ~20% overlap, and am processing with a bad pixel map, master dark/flat/bias. The individual panels are coming out OK enough.
I've tried two methods to process this
- process all the subs in one run as per some of the videos
- process each panel, then load each integrated
I either get like a "rainbow" or I get a stitched together image where the blending is horrible. Neither are an acceptable.
In the attachments I have where I've processed all six panels, didn't any further post processing (remove LP etc) and just save as a jpg and then put them on the screen in the right order and taken a screen shot. This is how APP should be processing the image I think. I've also attached sample images of the rainbow effect as well as two examples of the image where at least it processed, but the blending is horrible.
I had LNC with multiple iterations, scale moved to 1-10 from the default, MBB at 20%, registration mode set to mosaic etc. These are all settings I've seen in the videos and tried to follow. I also set it to 2000 stars in analyse stars. The subs are pretty clean as they are from a darker sky site too. I've been trying various different options for a couple of weeks now and am making no progress with having an even passable image.
Can anyone help with what is going wrong here please?
Thank you, Craig
Hi Craig,
For you rainbow issue, try to click the OpenGL button.
To improve the merging of your different tiles, you can also try :
- set 3000 stars in step 3
- uncheck dynamic distortion and same camera/lens in step 4
- in step 5, what mode has been selected, regular or advanced ?
Maybe some informations there : https://www.astropixelprocessor.com/community/main-forum/some-questions-for-creating-a-mosaic/#post-5805
Cheers,
David
Thanks for these tips David.
I had stars set at 2000, not 3000. I did have dynamic distortion and same camera enabled from memory and will go back and uncheck those. For step 5, I set this to advanced from regular because I saw Mabula do this in one of his video's. I'll go back and read that URL you posted too, thanks for that.
I'll report back with the result. I hope this works as one of the reasons I got into Astrophotogrpahy was to do mosaics as well and to date APP has not produced a decent result. Hoping these setting tips do the trick.
Thanks again.
One other quick point, I had scale start/stop set to 1-10, not 1-5 as per the default. I saw that setting applied in a video on doing mosaic's as well. Does that also sound right?
OK, I have just re-processed this using the above recommendations. I used the integrated six panels, rather than re-process the individual subs.
I am still getting the "rainbow" as the processing outcome. I have no idea why APP is doing this.
Does the rainbow is still there after clicking the opengl button?
I had the same problem and solved it by desactivating opengl.
No, this did not seem to make any different in relation to the rainbow.
I have gone back and tried to re-process all the 150 subs (6 x 25subs per panel) with the suggested settings mentioned above. This is taking, literally, days and it still hasn't finished yet. I am making no progress here by the looks of it. Wondering if this is a Mac thing or not? Because all the video's I see its using Windows and this Mosaic tool seems easy. I've had no success with any of the Mosaic's I've attempted with APP, in spite of getting OK results with single images.
I've emailed support at AstroPixelProcessor too back on the 27th of May and haven't had any reply yet either. I'll update this thread if I do get anything that can help with processing this.
I use APP with a macbook air from 2016, and it works fine (maybe a bit slow with 8go RAM).
If your ok, you can send me (PM) your 6 integrated files via wetransfer, I can have a look and try.
Here is the link in WeTransfer to the 6 panels https://we.tl/t-x5HJCWTEYq if anyone wants to have a look and let me know what I've been doing wrong that'd be great!!!
Each panel has been
- stacked (25subs each panel) with a bad pixel map, master dark/flat/bias. Calibration frames were all created with APP as well and stacked in APP
- no remove LP / calibrate background etc etc, they are just stacked and left in FITS format
I did get it to process last night using these updated settings, but was then unable to do any remove LP, calibrate background. The image was entirely washed out and I couldn't bring it back when trying to apply removeLP to the combined image. But at least there was some progress to get a stitched image. The panel's weren't well integrated though colour wise, but at least there was improvement on the smudging between panels.
Hi Craig,
When I load the six panels in APP and start a mosaic registration, APP requests the focal length of the telescope and the pixel size of the camera. Can you provide those please?
Thanks, Wouter
My apologies, again, for the lack of support lately. Mabula will inform everyone as for the reasons why.
Thanks for sharing the data, that might give us a better idea. Also, what software did you use to capture the mosaic and what was the selected overlap in % (edit: sorry, 20% as you mentioned). It might be good for us to see a few of your calibration darks/flats as well.
It shouldn't matter on which platform this is done (the mosaic process itself), the rainbow effect might have something to do with OpenGL and Mabula is at least working on getting APP to work way better with GPU support and more recent graphics libraries.
I downloaded the data and am looking at the separate panels. I do see a lot of color issues it seems in the background, could you upload like 10 light frames from 1 panel and your masterdark and masterflat? Besides that the stars themselves seem a bit strange as well, there's definitely distortion in them and something is really off if you look at the bottom right corner of panel 4 for instance. I'm curious if this is also the case in the separate frames. If so, something might be wrong with your optics which can cause troubles with nicely aligning the frames. Even though APP still seems to be able to do it given your further results. I'll now try the mosaic myself as well.
And another observation; in every frame there is a circle visible in the middle. This indicates you also have light pollution causing some issues. So to get towards better results, these are things that need to be solved first.
So this end result is still pretty bad, but at least the frames were registered properly without artefacts as far as I can see. The differences in the background are still big as that problem lies within the data and can't be solved without some serious work. I tried to control it in the seperate frames, but it didn't help too much, even made the middle of the nebula a bit too bright (I think my normalization went wrong here).
I used a scale start of 4, scale stop of 12, distortion margin of 3. These are big numbers, but I figured it might be necessary given the visible distortion. In the integration tab I used 5% MBB, 1st degree LNC with 4 iterations.
I captured this with a EdgeHD 8" with Hyperstar, so its a 400mm focal length setup at f/2 and these are 60 second subs. Capture software is Kstars/EKOS and the camera is a ZWO ASI 1600MCPro cooled to -10 with a gain of 139 (unity), so a one shot colour camera.
That last image there is a drastic improvement in what i have been able to achieve. Can I ask what settings you used to get that with? That's clearly where I need to be as a base position, you've done a far better job there.
I'll upload some of the lights and the masters I used as well shortly. Thank you so much for getting back to me here, this has been hard to make some progress on this and its great to get some help with it.
I posted some of the setting above the photo. I still didn't like the result in the sense of some over-correction, yours are far better exposure-wise so also in my processing something went a bit wrong. You might try the mosaic with those settings, then try to clean up the result. But there seems to be some reflection in your setup which shows as these center rings. This might also affect a proper calibration (not sure). Such a fast system can be very tricky (though it's awesome to have).
edit: You might even lower the exposure a bit as this target is pretty bright to begin with for F2. And nice to see a fellow EKOS user (I use both EKOS and SGP whichever I feel like trying 🙂 ).
On the light pollution comment, I have not applied any remove LP on any of those panels, they are just stacked and integrated. I'm uploading a full panel's lights as well as the calibration frames and will post that back ASAP once its been uploaded.
With the LP I meant reflection, as in a problem with light pollution inside your optic-train (or so it seems). The LP correction in APP accounts for LP from city light for instance, causing a gradient in the sky. This however if of a different kind that the LP correction can't solve. Just to bring that to your attention as something you might want to look into.
Thanks for uploading the other data!
OK, here is a 2nd link of a single panel's lights as well as the calibration frames I used in APP to process them https://we.tl/t-2l86gZtSdP
Downloading them now, will have a look at them tomorrow (camp-site wifi 😉 ). Thanks.
I am interested in the settings you used above though, as it looks far better than what I've been getting. Did you document what you set in steps 4,5,6 for that result by chance?
To one of the earlier questions on my setup - I think this was everything you requested
- focal length is 400mm
- full details on the camera I'm using is here - https://astronomy-imaging-camera.com/product/asi1600mc-cool
Camera technical details
Sensor: 4/3″ CMOS
Diagonal: 21.9mm
Resolution: 16Mega Pixels 4656×3520
Pixel Size: 3.8µm
Bayer Pattern: GRBG
Max FPS at full resolution :23FPS
Shutter: Rolling shutter
Exposure Range: 32µs-2000s
Read Noise: 1.2e @30db gain
QE peak: 60%
Full well: 20ke
ADC:12bit
Tab 4:
Scale start of 4, scale stop of 12, distortion margin of 3.
Tab 5:
Advanced mode
Tab 6:
5% MBB, 1st degree LNC with 4 iterations.
Not saying these settings are the best, I used quite large numbers for distortion etc, so maybe you can set those back, but that's a bit trial-and-error. I also worked on the panels a bit, before doing the combination. So I removed light pollution (which isn't really LP in your data, but it still helps) and calibrated the colors.
I've gone back using these settings to try and re-process this mosaic. So all subs, including calibration frames, from the beginning, to try and get a basic result. It has been running the registration for well over a day now and it is still only at 38%, just in the registration phase. I've given APP (v1.074.1) 12gig of RAM and all four core's and the machine isn't running much else. It has done this previously where I left it running for 3+days during registration and it still did not complete this phase.
Is it typical for APP to take this long during a mosaic.
With that many frames it might take a long time yes (based on the settings as well), that's why Mabula is focusing on getting GPU acceleration to work. However, you could also take parts of these files and integrate those, basically chopping it up in batches of 40 for instance and then integrate those together into a panel. Then you can mosaic together the main panels, it might be faster. I would process each sub-integration by trying to clean it up a bit as well (so removing gradients and such).
I was travelling for work so I decided to start a re-preocess of the 150 subs. After about ~9 days of it running continuously, it never even finished the registration stage. To me, six panels of only 150 subs shouldn't take this long to process.
I'm still no closer to getting a basic result for mosaic processing with APP.
I was travelling for work so I decided to start a re-preocess of the 150 subs. After about ~9 days of it running continuously, it never even finished the registration stage. To me, six panels of only 150 subs shouldn't take this long to process.
I'm still no closer to getting a basic result for mosaic processing with APP.
I just tried to process just the six panels that were processed individually and I get the rainbow (see the attached screen shot)
Tab 4:
scale start 4, stop 12, distortion market 0.03, registration mode: mosaic
Tab 5:
mode: advanced
Tab 6:
weights: equal; integrate: average; mode: full; 1st degree LNC, four iterations, MBB at 5%. no outlier rejection.
I have no idea why this isn't even producing even a basic result at this point in time.
I Craig,
Sorry, I was quite busy lately, you know, work, family stuff with young children 😉
Could you please try this : before loading your mosaic, click the openGL button and then show us the result (the button should be in orange and not green). I had the same issue (macbook air from 2016), and still has it : if the mosaic is more than 3 or 4 panels, I have a rainbow displayed.
@mabula-admin advised me to click the OpenGL button, and then the mosaic was properly displayed, but with a much slower APP.