Pixinsight image integration problem

[ChrisM]

I've used PI occasionally over a few years and never had an alignment issue before.

Now when I try to register a number of images then use image alignment to combine then, the resulting image trails the stars with no apparent attempt to align them.

Whilst the star alignment process reports that alignment was successful on all images, I suspect that the issue is with this process and not the image integration.

In the star alignment panel, I have selected a reference image and set the working mode to Register / Match Images. There are plenty of stars to allow star detection to work properly.

I can't figure out what I'm doing wrong - so would appreciate any help please. I am using Version 01.08.05.1352 Ripley (x64).

[Atmos (Colin)]

Mono or OSC?

[ChrisM]

Originally Posted by Atmos

Mono or OSC?

OSC - same as I usually use.

I've done some more investigation and am convinced it's the registration process. None of the registered images look any different to their respective originals.

[RickS (Rick)]

Are you sure that you're using the files that were output by the StarAlignment process? I just helped someone on the PI Forum with the same issue. It turned out that the registered files were going to a different folder from the one he expected.

I've never heard of StarAlignment failing to register images without displaying error messages. It would be worth checking through the process console log carefully. Another thing you could try is opening a few of the files and applying StarAlignment to them as views and watch what happens.

Cheers,
Rick.

[Atmos (Colin)]

Are you trying to register debayered frames? The StarAlignment does not deal well with frames that are debayered (mono) as opposed to bayered (RGB).

[ChrisM]

Originally Posted by Atmos

Are you trying to register debayered frames? The StarAlignment does not deal well with frames that are debayered (mono) as opposed to bayered (RGB).

Thanks Colin - the RAW images I'm using are straight off my Nikon DSLR - the same camera source that I've used a few times previously.

[ChrisM]

Originally Posted by RickS

Are you sure that you're using the files that were output by the StarAlignment process? I just helped someone on the PI Forum with the same issue. It turned out that the registered files were going to a different folder from the one he expected.

I've never heard of StarAlignment failing to register images without displaying error messages. It would be worth checking through the process console log carefully. Another thing you could try is opening a few of the files and applying StarAlignment to them as views and watch what happens.

Cheers,
Rick.

Thanks for the suggestions Rick.

The files produced by the alignment process have the '_r' appendix on them and I've pointed the integration tool to those files so am confident that that is not the source of my problem.

I had a look back over some previous projects and viewed the registered files (that I had saved). There seemed to be some variance with how these files were processed, in that for some projects, the registered files didn't look to be modified, whereas in other projects, the registered files had obviously been translated (eg. rotated slightly) ready for the integration tool to superimpose them. I integrated one set of registered files and that worked just fine.

I have noticed that the star alignment tool is layed out a little differently to how it was a few years ago when I first started, and also has a few different check boxes, so I am wondering if I've got those all set appropriately.

The other thing I noticed, which is probably highly relevant, is in relation to what the process console displays for each file (I had 60 in this batch). Because I didn't use any guiding (just tracking), there is a degree of drift, and this target was close to the celestial equator so the drift was noticeable across the hour or so during which the images were exposed (60 x 1 min subs). Therefore (and as in the past), I would have expected to see the translation matrices with progressively more and more offsets (ie. higher pixel adjustments) relative to the reference image - which I selected to be the first one that I took in the whole sequence. If I recall correctly, it would not be uncommon to see the pixel adjustment increase by say a few tens of pixels for each exposure. This was especially so when I did in-camera darks, but I didn't do that on this occasion.

I've just run the star alignment tool again, but to speed up the process I selected every 10th image. I've copied the process console workings to a Word file for easier viewing, and could send that to you Rick if you felt inclined to have a closer look. From the workings, there seems to be some erratic variations in pixel offsets. It is as though the reference file isn't being referred to properly, although I did include the reference file in the target image list, and it came through the process with a 'clean bill of health' - ie. no adjustments whatsoever, which is exactly as it should be.

I have also just tried opening two of the files as views - and that made no difference either. The new file was identical to the original RAW file. For this test I selected the reference image to be the first of the batch of 60 that I took, and the file to be aligned was the last of the 60 (ie. I was looking for maximum effect).

So, something is not right and I'm at a bit of a loss as to what to try next.

Cheers, Chris

[troypiggo (Troy)]

G'day Chris

Have you tried with the BatchPreProcessing script to see if that at least gives better result in terms of alignment? Might help highlight where you're going differently?

Other suggestion would be to post some screenshots of each process settings, and maybe a few of the files, at least the one you're using for registration?

Another thought - are you sure the newer trial "_r" suffixed files are in fact newer. ie you haven't changed a setting where you're still actually looking at original poorly aligned subs, but you think you're looking at the output of further testing?

Sorry if you're tried the above, I'm not trying to tell you how to "suck eggs". Just some thoughts I'd try if it were me.

[RickS (Rick)]

Hi Chris,

The default StarAlignment settings generally work for me and I've never had to tweak the basic alignment parameters to get a good alignment. I have heard of cases where StarAlignment was confused by hot pixels, so I guess that's a possibility.

Can you upload your reference and a small number of subs to Dropbox or somewhere similar? I'd be happy to have a play and see if I can reproduce the problem and then look for a way to correct it. Also happy to dig through the process console log and see if there are any hints in there.

I'll send you a PM...

Cheers,
Rick.

[ChrisM]

Originally Posted by troypiggo

G'day Chris

Have you tried with the BatchPreProcessing script to see if that at least gives better result in terms of alignment? Might help highlight where you're going differently?

Other suggestion would be to post some screenshots of each process settings, and maybe a few of the files, at least the one you're using for registration?

Another thought - are you sure the newer trial "_r" suffixed files are in fact newer. ie you haven't changed a setting where you're still actually looking at original poorly aligned subs, but you think you're looking at the output of further testing?

Sorry if you're tried the above, I'm not trying to tell you how to "suck eggs". Just some thoughts I'd try if it were me.

Thanks too Troy. I either deleted the older files first, or overwrote them to ensure the integrity of each new test.

I had a very quick look at Harry's Astroshed last night to see if he had updated his video tutes to match the PI updates, and did see the batch processing example. I am yet to try that but will give it a go.

In case it helps, I've attached a copy of the processing console info for the example where I selected every 10th pic in my sequence of 60 images. The first pic (9480) is also the reference image. When I compare the first and last raw images, the drift across the field of view is of the order of about 3%, so definitely a requirement for proper registration.

I'm probably not going to be able to get to my latop for a day or so, so may be delayed in responding to any replies.

Cheers, and thanks,
Chris

[RickS (Rick)]

Originally Posted by ChrisM

In case it helps, I've attached a copy of the processing console info for the example where I selected every 10th pic in my sequence of 60 images.

Some of the translation deltas are suspiciously small.

You could try bumping up a few of the star detection parameters: Noise Scales, Hot Pixel Removal, Noise Reduction. That will help if the star detection is being confused by hot pixels or noise.

Cheers,
Rick.

[ChrisM]

Originally Posted by RickS

Some of the translation deltas are suspiciously small.

You could try bumping up a few of the star detection parameters: Noise Scales, Hot Pixel Removal, Noise Reduction. That will help if the star detection is being confused by hot pixels or noise.

Cheers,
Rick.

Thanks Rick - I don't have time to try that now, but will PM you some sample files to have a play with.

Cheers, Chris

[troypiggo (Troy)]

From the console output, it looks like you're shooting at extremely wide angle - 15mm lens? That may explain the very small translation deltas.

[RickS (Rick)]

Hi Chris,

Thanks for the test subs. I was able to reproduce the problem. It looks like StarAlignment is confusing hot pixels in your data for stars. Increasing the "Hot pixel removal" parameter on the Star Detection tab from 1 to 2 was enough to get a good alignment on the 4 test subs you gave me. Give it a try on the full data set and let us know how it went. You could crank it up a little more if you still have some subs that aren't registering.

I normally use CosmeticCorrection as part of my calibration process so there aren't usually many hot pixels left in the subs that I register. Your subs don't look particularly bad, but apparently there are enough hot pixels there to confuse StarAlignment with the default parameters.

Cheers,
Rick.

[sil (Steve)]

Reading the thread it sounds like your subs aren't too noisy, just hot pixels being confused for stars. What I often do in troublesome cases is to use Dynamic Alignment to nicely align by hand three or five frames myself then put those save frames into Image Integration. ? This gives me a file I call tmpIntegration.

THIS is what I use as my integration target. Hot pixels and strong noise are greatly reduced in this file so confusion for actual stars goes away quickly, Confidence in found stars is increased and registration using this target is always good.

It also saves time mucking about with alignment settings I've already tweaked for the bulk of my captured data.

[ChrisM]

Originally Posted by RickS

Hi Chris,

Thanks for the test subs. I was able to reproduce the problem. It looks like StarAlignment is confusing hot pixels in your data for stars. Increasing the "Hot pixel removal" parameter on the Star Detection tab from 1 to 2 was enough to get a good alignment on the 4 test subs you gave me. Give it a try on the full data set and let us know how it went. You could crank it up a little more if you still have some subs that aren't registering.

I normally use CosmeticCorrection as part of my calibration process so there aren't usually many hot pixels left in the subs that I register. Your subs don't look particularly bad, but apparently there are enough hot pixels there to confuse StarAlignment with the default parameters.

Cheers,
Rick.

Success! Thanks Rick - the setting of 2 for hot pixel removal seemed to work, although I suspect that the High Rejection image may contain some data, as it has dozens of parallel coloured streaks (like drifting stars) plus fixed points where the brighter stars are located.

When I get more time I'll have a look at CosmeticCorrection, and will also check out Dynamic Alignment as suggested by SIL.

In the meantime, I'll further process what I have and see whether I can identify the Mag 19 star that was my target.....

Cheers, Chris

[RickS (Rick)]

Originally Posted by ChrisM

Success! Thanks Rick - the setting of 2 for hot pixel removal seemed to work, although I suspect that the High Rejection image may contain some data, as it has dozens of parallel coloured streaks (like drifting stars) plus fixed points where the brighter stars are located.

Glad to hear that it helped, Chris.

If you have stuff that looks like data in your rejection maps that's usually a sign that your rejection parameters are too restrictive and should be tweaked. The ImageIntegration defaults are much too strict in my experience.

Cheers,
Rick.

[ChrisM]

Originally Posted by RickS

Glad to hear that it helped, Chris.

If you have stuff that looks like data in your rejection maps that's usually a sign that your rejection parameters are too restrictive and should be tweaked. The ImageIntegration defaults are much too strict in my experience.

Cheers,
Rick.

I had a play with CosmeticCorrection and Dynamic Alignment last night, but would need to experiment further to properly test them.

Also, I just realised why I probably had the Star Alignment issue this time and not previously - it's because I didn't use in-camera noise reduction on this batch. Might teach me a lesson!

Cheers, Chris