Next project: bat and squid thoughts/planning

[ram]

Hi all, I've started on my next capture, which is going to be the Squid nebula (Ou4) primarily followed by a widefield of the Bat (Sh2-129). The former resides within the latter. My Squid capture with the FC100 captures a lot of the Bat, I'd say at least about 50% but perhaps as much as 50%. Nonetheless, I plan to do it both ways and ideally integrate the higher resolution Squid data (in theory) with the wide field data without needing the same amount of exposures.

I've already captured about 20 hours of data on this target, started with S2 and Ha since the moon was up a bit (5 hours each, which may be enough) and then now I'm going all in for the O3 which is the most important part since the Squid is all in O3 and it is VERY VERY faint. I'm at 10 hours and can see the outline but not the internal structure. I'm also dealing with some poor frames (clouds) so I think this may pay off in the end to have a clean data set. Some of the best images I've seen of this target capture about 30-40 hours on O3 alone.

Here's the image in case people are curious: this is by someone else at AstroBin: https://apod.nasa.gov/apod/ap161027.html

My question is what should I look out for in attempting to do this? Does it make sense when I start processing my SV70T capture to use Drizzle to blow up the size of the image so I can better "fit" the Squid (captured with the FC100DF) within the Bat? Right now as I said, I'm getting the "bowl" the Squid sits in. This is what my current framing sort of kind of looks like: http://www.starpixel.it/en/galleria-2/nebulose/sh2-129/ - so you can see between the top view and bottom view, there's different aspects to it. I could do a mosaic but I think this strategy may be the better one if I can get it to work. As long as I can bridge the seams when I put in the higher res image into the lower res one, it should be okay. So should I be thinking of drizzle integration or will it work without or if anyone has tried things like this and has advice to offer, please do so. I'd be eternally appreciative.

Also the object is right near the pole when I start so there's some movement in my mount over many hours (which is usually rock solid otherwise). Over five hours, there's about a hundred pixels being traversed (usually it's about ten or less and this is entirely unguided).

Thoughts?

--Ram

[STEVE333]

Hi Ram - I don't think this will be difficult.

1) Take the S2, Ha, O3 data.
2) Open StarAlignment and, with the Ha image as the Reference image, align the S2 image to the Ha image.
3) Open StarAlignment and again set the Ha image as the Reference image.
3a) Set the O3 image as the Target Image.
3b) Drag the triangle (Apply icon) onto the O3 image. This will rescale and align the O3 image to the Ha image.
4) Combine the SHO data as you normally would.

I believe that sequence will do what you want to accomplish.

Steve

[ram]

Hi Steve, thanks - yeah, I think that would work but I'm wondering if a simple combine is the optimal way of doing it especially to preserve the higher resolution of the inlined image. There are two sets of Ha, O3, and S2 data taken a different resolutions, one of which is correctly sampled for my seeing and the other one is undersampled (the lower resolution one). In effect, they are different images with overlapping framing of the same object. Because it takes 20-40 hours of time on the Squid (central) part, I would prefer to avoid redoing that and use the older data at preserving it at the higher resolution and cut the time on the second image and I also would like the integration to happen seamlessly (i.e., without seeing a seam which you can actually find in the IC1805 Heart image which did the same thing you describe above if you look VERY VERY closely and this is after I clone stamped out the obvious places where you could see the seam). The seam question isn't what I'm really asking though (which has to do with matching the backgrounds, etc. basically linear fit may work or I may just get lucky as I did in the IC1805 case), I'm thinking about the resolution change.

The Squid data is captured at higher resolution with an image scale 1.61"/px (which is supposed to be "properly sampled" according to the CCD calculator on most seeing conditions: https://astronomy.tools/calculators/ccd_suitability), and that is to be combined with the Bat data captured at a lower resolution of 2.33"/px (which supposed to be slightly undersampled according to the same calculator for my conditions).

The basic question I have is whether drizzle integration of the undersampled image and then combination will help preserve the higher resolution Squid data or if there's a better method or if drizzle isn't necessary (but I don't see how without drizzle I can preserve the higher resolution of the Squid data when I rescale, i.e. resize, which in this case would get it down to the lower resolution of the widefield image).

There's probably a way to scale up the drizzle properly so that the final sizes of the two sets of data are close (i.e., perhaps the 2.33"/px can be drizzled x 3 and then the other one can be drizzled x 2 which would put them on an even footing in terms of resolution? it's been a while since I did drizzle and I am going by memory) and then doing a combine like what you describe should do the trick.

--Ram

[STEVE333]

I think I now understand your dilemma. I hadn't realized that you would have two complete separate sets of Ha, O3 and S2 data.

I'm not that familiar with drizzling, but, I get the idea you are suggesting. Because drizzling does introduce noise (more resolution brings with it more noise) retaining the resolution for the Squid may be at the cost of more noise in the Bat. However, you may have enough time on target that the tradeoff is still worth it.

If you drizzle the Bat and then "star align" the Squid to it, you will then have matching images to combine while having retained the resolution of the Squid.

You may already know how to do this, but, if you don't, I have a simple technique for inserting a portion of one image into a second image. It also allows you to adjust the background level/color of the inserted image to match the background of the "master" image. Just FYI.

Good luck with this project. It's a tough target, and, I look forward to seeing your results.

Steve

[STEVE333]

Ram - I thought I would demonstrate the inserting of a region from one image into the same region on a second image. The image below shows what it looks like when I took the inside of the heart nebula from your second (greenish) image and inserted it into your first image. I don't believe you can see any indications that the two images were combined. Notice also that the Fish Head nebula doesn't have any green because only the interior of the Heart nebula was inserted.

Heart Blend by STEPHEN KING, on Flickr

Hope this makes sense.

Steve

[ram]

Wow, that's really cool - I was dealing with this for the S2 data here as well. I don't have any fixed method - I either register the whole thing and hope the backgrounds match (worked one time), use linearfit (worked another time), or discard the data (once).

So I'd appreciate the steps you took to create a seamless image with matching backgrounds. I will definitely have to do this for this project... thanks Steve!

--Ram

[STEVE333]

Ram - I created a tutorial describing how to do your "blending" of the two images. It is at the link below:

viewtopic.php?f=22&t=12677

Hope this helps,

Steve

[ram]

Thanks Steve! I think your use of the GAME script is very clever/innovative but I wanted to tell you what really helped me out the last time you showed me how to clean out my corner amp glow. It's not just getting the shape you want to combine (which GAME allows for to an exquisite level of detail but sometimes I find range selection may be enough) but really how the border regions are handled so there's no seam. I see once again you use Convolution to achieve this.

As you know, I previous didn't use masks much and I was new to it. The last time I did masks years ago extensively was in a PS like environment where I could create whatever mask I wanted by painting. But this wasn't possible with PI easily but I was able to use range selection along with PixelMath and CloneStamp to create the corner range I wanted to mask out but it had a hard border which is why all my previous attempts to remove the corner amp glow sucked. The border would get visible if I went too far and I always struggled to not show the border, etc.

The main trick I didn't realise was the notion of making the borders smooth using Convolution: that really smooths out the borders and makes masks really powerful. So thanks a lot for that!

--Ram

[ram]

In addition to the posting I made on clouds (about 10% of my frames are with partial clouds, probably should throw them out but there's faint signal there), there are other issues with this target I wanted to bring up. In particular, I have to deal with star halos caused in part by the cloudy data (anyone how to use the StarHaloReducer script?).

At 30 hours total, I have managed to bring out some detail and it doesn't look too bad. In particular, LHE is working for the first time to enhance feature contrast. Here's a sample image after DBE, CT, some NR (without a mask, since I wanted to see how LHE would work on the NR data) and LHE (bright, I could tone it down but I wanted to see what it would bring out):

And here's the version with just DBE and CT:

I've not done any deconvolution yet or any of the usual preprocessing and the corner brightness due to amp glow need to be masked out (image is medium quality JPEG rotated 180). I was just worried that I wasn't able to pull details out with just the straight integration so I wanted to see what would work. I surprisingly was able to do a good DBE this time and CT was enough to bring out some detail but my previous attempts weren't that good which is why I wanted to try LHE; I've not had much luck with it but I was desperate. This isn't a case for HDRMT or DSE so that leaves us only with CT and LHE to enhance feature contrast.

People I think mask out the surroundings other than the squid but you can see other than the amp glow there is some very faint O3 data - probably gases blowing away or something.

For this target, the ideal situation would be to image about 80 hours of only O3 data and then throw out like 40 hours worth to keep the best frames. Or perhaps a 50-40 split will do. Definitely keeping the bad frames as usual seems to hurt this the most because it's so faint and also there are bright stars ionising the O3 signal. The clouds are making these bright stars look fuzzy.

I think I will do about 40+ for this O3 with the FC100DF and then do another 20-40 on the SV70T which should get to my goal of getting a lot of data for each filter at different resolutions on this target and keep revisiting it. For Ha and S2, I think 5 hours is enough but I may just double that, so get like 20 hours total there. But we'll see how the weather cooperates.

After this and the Lobster and the Jellyfish for which I have partial data I think I may be done with this season but we'll see. The Jellyfish is another target for which if this data combination stuff works out I can pursue further to create both mediumfield and widefield images.

--Ram

[ram]

One more thing: I will do flats at the end (perhaps I should do it in the middle?) and try it out to remove any artifacts and even out the light gradients. It doesn't always work for me but I will try it. Given the imaging done over many nights near the pole (mount moves more) there's none of the dust bunny stuff on there. Flats will only be useful for evening out the light gradients.

Also while I do have halo stars everywhere, my mask for brightening the nebula isn't perfect. I just created it from the first set of frames (about 5-6 hours worth) and didn't subtract the stars so LHE was applied to the stars also which makes them brighter. So that's one thing to keep in mind once I collect enough data to say "enough" and then redo the deconvolution mask which will be useful for LHE and other things.

--Ram

[ram]

As I can't post new images since this project is going to take up at least another moon cycle, this is what I've got thus far, yet another fake ass teaser:

I have collected all the data for the narrowfield image which I'm processing now. Testing out PI 1.8.8-6 which has StarNet in it and so creating a starless image is now a one click process except that I'm not sure how to get rid of the big star and the other small ones around it that don't get removed. I think with Steve's method I had more control over that so I'll try it both ways and see - I'm still working through pre-processing stuff with PI's new version there's this Adaptive Normalization that might be worth trying and so on (all this is with 1.8.8-5 - just the SN part is the new version). This is just a simply lazily done image (~53 hours) that was then rendered starless, basically the second image above, and then fastly processed. The Squid really stands out I think with the starless technique. Given how discrete this object it, inserting it back into other fields (either the Bat "bowl" or the full Bat or a RGB background) should be easy enough.

I was plagued with a bunch of flexture issues in both my OTA and my camera (which I thought was due to the mount) so given that, I think it came out okay but I'm glad I'm getting a chance to redo with the SV70T - the combined data should at the very least make it robust (the target is so faint that going from 25 hours to 53 hours didn't do much, given the effort involved). I can see compared to other people's images I'm having some side detail pop out which is pretty cool.

My plan is to shoot some OSC data of this field and then then combine that with the Squid by itself. I was just going to do a mono Squid (as well as the narrowfield Squid - I have another ~25 hours of Ha and S2 data of the Bat "bowl" and I've started the widefield that includes the full Bat) or a mono Squid with SHO stars but I saw someone do this with RGB stars and I thought it was interesting (i.e., Squid by itself as I'm showing without the Bat background but with the RGB stars).

Since my spacers are stuck to the connectors on the Tak FC100DF which I used to the above image, I can't mount my OSC on it yet and I don't want to move the camera on the SV70T since it is tilted at exactly the right way (around a 45 degree angle) to get the framing I want.

--Ram

[sdbodin]

Fabulous Squid, but even 25 hours is more than my ADD can stand.

Steve

[ram]

Thanks! ADD?

If it means what I think it means, yeah, I've been tempted to scrap the whole thing a few times (and even quit AP, not because of the target but because of the issues that came up, I was so worried I had botched the lubrication of the MyT this summer but the mount is behaving really well but there were other flexture issues). I wanted to switch to the Lobster, get back to something else, etc. When do you these long runs, you really realise all the problems you gloss over just by picking the right target, right framing, right timing, etc. Doing this very long run in one season requires some planning (my total is about 80+ hours and I'll easily cross 100 on this but I hope to get at least 3-4 distinct images out of all this data with different perspectives and framing).

For instance, my ideal imaging time is between 11/midnight to 6am but this target now goes behind trees now at 430am (but a month ago, it was around 530am) and I think it just gets within 2 hours of the meridian (so I don't have to do a flip but just image it upside down) around 1030pm now at my location (but again a month ago it was perfect starting around midnight). If I'm going to put in another month on this, then that's shifting everything and cutting into my imaging hours... something I didn't think about when I started. But I'm determined to get this done... in some ways, it's easy to for me to spend about 1-2 hours each clear night (optimally it's ~1 hour for both setup and take down) doing the same thing than 2-4 hours fiddling around wit a new target, framing, debugging, etc. Always something or the other. I was going to write a long post about it but then I decided against it since I realise I probably am not the only one (there was someone who posted here "there's always something else" or something like that in similar frustration which echoed my views on this hobby).

--Ram

[sdbodin]

I understand, but back in my day there were no fancy acronyms for just not having the patience for it.

Steve

[STEVE333]

The Squid is looking wonderful Ram with lots of internal detail showing up. Really watching this project with interest.

I recently learned about the MureDenoise script. I've been playing with it and find it does an excellent job of "blending" regions that have speckled low-level nebulosity (like the interior of your Squid). Do you use MureDenoise?

Steve

[ram]

I'm nervous about the pulling the whole thing off so I tend to "peek ahead" and play around with options. Specifically I've only collected ~3 hours of Ha widefield data of the Bat itself and I wanted to see if everything was working out right (especially since I changed the framing to a 45 degree angle between the narrowfield and widefield images) and if I could merge things properly. The idea is that I'd merge the three narrowfield to widefield masters for each of the three filters and then combine and process normally. I also wanted to use some of the fancy merging methods Steve posted but just getting the framing right before that can be done is what I was looking into.

This didn't turn out well in terms of brightness, etc. but there's really zero processing; I pushed the lightness on the three hour Ha data so when I PixelMathed the O3 weak signal Squid, the rescaling causes it appear washed out but overall one gets the idea. You can see where the frame borders are but even without any fancy combination it combined reasonably well I thought. This doesn't include the other Ha and S2 data but they'd just overlay on top of the Ha signal though they'd only cover the part in the inset frame (the borders are just out to the triangle you can see past the middle to the right) where it is already bright anyway - it's the fainter bits on the other side I'd need to spend more time on. I'd likely also get some widefield O3 data (at least one night's worth if not more) so I can drizzle and combine.

Thanks for looking/commenting!

--Ram

[ram]

The Squid is looking wonderful Ram with lots of internal detail showing up. Really watching this project with interest.

I recently learned about the MureDenoise script. I've been playing with it and find it does an excellent job of "blending" regions that have speckled low-level nebulosity (like the interior of your Squid). Do you use MureDenoise?

Steve

Thank you Steve! Lots to do still... so I'm doing little bit here and there. I'm just redoing a lot now with version 1.8.8-6 since Juan said there were some significant changes and updates.

No, I don't. I come across it here and there but once in a while I run into some other caveats but people who love it, love it. I'm following Jon Rista's tutorial now (or trying to) for NR and he doesn't get into it but if you have a good guide or good advice please let me know. Especially as I posed one form of the blended image (which I just posted above - the fully widefield). So are you saying MureDenoise is good for blending regions like the Bat and the Squid both of which have weird nebulosity and the Bat nebulosity hurts seeing the the Squid which is why it was only recently discovered I think? Or just one of them need to have low level nebulosity?

As I mentioned, I plan to have one image of just the Squid only with an RGB background if possible* and then other images with the Bat bowl (the center part or what I'm calling the narrowfield) and the whole Bat widefield. I imagine I don't have to take too many exposures with my OSC or just the stars - probably should do it with the SV70T since it'll be smaller stars that can be scaled up if needed, so halos, etc. would be better contained...

BTW, about StarNet+ did you see how I could remove the central star/group using PI's StarNet (you used to use StarNet before you moved to PI for star removal right)? Or use your star removal technique?

--Ram

[Graeme1858]

It's a pleasure to follow this thread and the level of expertise being discussed. Even if I don't fully understand it all I can aspire to!

Looking forward to seeing how your squid turns out Ram.

Regards

Graeme

[STEVE333]

No, I don't. I come across it here and there but once in a while I run into some other caveats but people who love it, love it. I'm following Jon Rista's tutorial now (or trying to) for NR and he doesn't get into it but if you have a good guide or good advice please let me know. Especially as I posed one form of the blended image (which I just posted above - the fully widefield). So are you saying MureDenoise is good for blending regions like the Bat and the Squid both of which have weird nebulosity and the Bat nebulosity hurts seeing the the Squid which is why it was only recently discovered I think? Or just one of them need to have low level nebulosity?

BTW, about StarNet+ did you see how I could remove the central star/group using PI's StarNet (you used to use StarNet before you moved to PI for star removal right)? Or use your star removal technique?

--Ram

MureDenoise is strictly a noise reduction process. When I mentioned "blending" I was referring to how it can take "low level "speckled nebulosity" and smooth it to out. The side-by-side comparison below shows what I mean. The image on the right is the original stacked image and the image on the left is the same image after it has been noise-reduced using MureDenoise. Notice how the "speckled nebulosity" has been smoothed out to a much more pleasing appearance. This was 2 hours of Sii data on the Wizard nebula, and, there isn't much signal. BTW if you click on the image below it will open up in Flickr to give an expanded view.

Mure Noise Reduction by STEPHEN KING, on Flickr

The link below is the only tutorial I have read on this so far. It was enough to get the results shown above.

https://astroimages.weebly.com/mure-denoise.html

One caution, MureDenoise must be applied BEFORE any other processing including ABE or DBE!


As for the removal, sometimes Starnet works best and sometimes the new star removal technique works best. I don;t know any way to choose beforehand.

The Squid data is looking awesome! Really having fun watching this one.

Steve

[ram]

Hi Steve, I tried out MureDenoise and it worked really well! The SNRWeight value just shot up - it was the first thing I did.

This is done without any masks right? It seems to be smoothing out the nebula really nicely but I think slightly adding some noise in the background. Is it not meant to be used on the background also?

--Ram