I think my images are going well until I try to color them in the color module

[Stefan B]

i'm still not sure how the restore function is supposed to work.

Have you already tried this?

viewtopic.php?f=4&t=2275&p=11146#p11146

Regards
Stefan

[xonefs]

i'm still not sure how the restore function is supposed to work.

Have you already tried this?

viewtopic.php?f=4&t=2275&p=11146#p11146

Regards
Stefan

I’m on mac so out of luck with that…

[Stefan B]

Nothing to do with OS. I think you looked at the post below the one I actually linked. Direct quote:

After applying NR and saving you can click 'Undo'. Then the 'Restore' isn't greyed out and you can go back.

My understanding is that tracking has to be active for Restore to be available.

I think the crucial point is that restore only works while tracking is active so the software can 'undo' the previous steps. After switching off tracking ST forgets all the steps and restore is greyed out.

[xonefs]

Nothing to do with OS. I think you looked at the post below the one I actually linked. Direct quote:

After applying NR and saving you can click 'Undo'. Then the 'Restore' isn't greyed out and you can go back.

My understanding is that tracking has to be active for Restore to be available.

I think the crucial point is that restore only works while tracking is active so the software can 'undo' the previous steps. After switching off tracking ST forgets all the steps and restore is greyed out.

Ah… I’ll try that. I had tried to activate tracking again and that didn’t work thought that's what you meant first time I saw that.

Edit: yes that worked thank you!

[xonefs]

So I had started working on another version to try to do entirely in pixinsight to compare using an old workflow

started no binning- crop then ez processing suite - decon on each channel, ez processing suite denoise on each channel. then combined to RGB, stretch all channels, extracted stretched luminance in pix.

At this point I remember how bad I am at pixinsight and how irritating it is and I don't know how to fix the contrast so I try a very bastardized workflow and import the stretched luminance to startools, run it through some contrast, sharpen, and superstructure. export the luminance and bring it back to pixinsight to recombine channels.

at this point I stop trying to fix the color from here as I realize that's more of a pain and I would need to try some different combinations first. however, ignoring the green color this still looks a lot better to me than anything I've been able to get out of just ST so far. There's just much better contrast around the edges of the nebula and it seems to have more depth (even though there is less contrast in center).

if the color was right I would be more happy with that and I really like the contrast/detail/presence it has and how it seems to glow and look natural

this is my last all ST attempt just now:

I have made significant improvements in my ST processing but I still am finding it lacking in contrast/detail around the edges and the center doesn't look as good/natural (maybe that's the slight HDR I just don't like). I'm not sure what needs to be done to get that kind of quality I'm going for starting in just ST or why I was able to get it looking better to me with my weird hybrid workflow.

[Mike in Rancho]

Nothing on my sample mimic? That bad huh? Well, I tried.

I also haven't had any luck yet with the new star shape generator in mask, though I've only experimented with the settings some. Seems to get way more than just some of the extended halos I was hoping it would, and instead looks like a full field jigsaw puzzle. Always just go back to one of the regular auto star masks and work on settings from there.

The restore often just gets you past the initial steps, such as composite, crop/bin, and maybe wipe. Ivo was noting the other day on CN that in the Linear, Wiped, Deconvolved option of restore, the deconvolution you made in the pre-restore session was actually taking into account all the other adjustments made first (like contrast, HDR, sharpen), so really might not be fully applicable unless those are then repeated. If I understood it right, that is. But restore to linear/wiped is still awful handy, or even original if wipe needs to be retried.

[xonefs]

^sorry I think that was between ivo’s comments and I missed it. No not that bad, but I don’t think anything i’ve seen yet from myself included out of ST that gets close enough to the look of those in terms of detail/contrast and overall look. My bastardized pix/ST workflow above seems closest to getting some of those qualities aside from the color so far.

I really want to be able to figure out how to get it out of just ST or even better exceed those.

[admin]

A couple of things;

I was slightly puzzled by some of the detail I noticed in the PI/ST hybrid.
It appears to me the hybrid version pretty much uses Ha only as a base for luminance, which will obviously yield/accentuate vastly different detail.

E.g. just quickly processing Ha with a standard workflow yields this, which appears (to me) to come close to the detail prevalence of the PI/ST hybrid;

Your expectations for a dataset (and comparison thereof to other images) should be greatly dependent on the amount of data you acquired for each band - an object will look different in different bands. The optimal signal blend (in terms of luminance signal) is dictated by that blend. Your exposure times should reflect your goals (e.g. the detail of the object you wish to bring out).

You don't need high exposure times for the bands whose detail you're not interested in; you can make due with much lower exposure times for the purpose of coloring, while you can spend more time acquiring the signal that you are truly interested in (which appears to be the Ha signal).

I further inspected the FITS headers and found that it shows very different exposure times (S-II 5400s, Ha 3600s and O-III 5220s), than what was specified in your processing logs. This matters a great deal when importing your datasets in the Compose module (and for decon results).

Weighed properly in the Compose module, you will find the fine detail recovered by decon (particularly SVDecon) superior to what you can recover from just the Ha-only set - decon just has much cleaner (more) signal to work with. At the same time, you can/should also expect the nebulosity to render differently, as your are showing a very different blend of bands with a higher emphasis on detail from the other bands.

So, if you like the detail of the Ha band most, you can just use the Ha for your luminance (set red/S-II and blue/O-III to 0 exposure time in the Compose module), and use the S-II and O-III for coloring only. It won't maximise your signal for the purpose of fine detail, but it sounds like you'd be happier with just using S-II and O-III for coloring. If so, for your next object, spend more time acquiring Ha in lieu of S-II and O-III.

Hope this helps!

[xonefs]

Thanks.

If it appears Ha was only used as lum that was not intentional and I'm not so sure. They were stretched and combined, then total luminance extracted. Ha could have still been more dominant and that could have happened unintentionally (I don't really know what I'm doing or how pixinsight extracts luminance in that situation, but I thought it would grab sum from all of them).

I can't really tell much of a difference where you used Ha as lum there from some of the other versions though honestly.

I did keep the exposure time equal in ST. I thought those times were more important for LRGB calculation and I wasn't sure how it would apply to narrowband. My reasoning for keeping them equal was that I thought since Ha was the shortest time that startools would make Ha even more dominant in the extracted luminance than it already is if I put in the actual times.

Is that wrong and not what would happen? I thought if it uses those to balance them, and if one had shorter times than the others that startools would try to boost the shorter channels to compensate their contribution. which would have the effect of the extracted luminance being even more mostly Ha. I was not aware it was also important to decon.

For RGB those times make sense as being important since each contributes directly to a proportional section of total luminance, but in narrowband Ha is always going to be hugely dominant and I don't think everyone always wants that in their final image.

Ha is almost always going to be the dominant signal, and generally needs much less exposure time to achieve similar SNR to oIII and SII, which is why I will often shoot a bit more SII and Oiii since they are weaker (and even shooting more of them here they are still not as clean as Ha).

I have heard that as a technique to focus on ha and then shorter on the others and just using ha as lum, but I've been disappointed in the past not gathering enough signal on the other channels, I generally process things weeks or months later so it's not easy to go back, and I want the flexibility to use all of them.

I don't really want to just use Ha as lum, and I think there is great detail in Sii I would want to bring out. I had thought about trying with just those two and keeping Oiii out of luminance however but haven't gotten to it yet.

[admin]

Thanks.

If it appears Ha was only used as lum that was not intentional and I'm not so sure. They were stretched and combined

Non-linear stretching and then combining makes no sense... Is it likely you just stretched until you saw the detail "you liked" (which is mostly visible in the Ha) and then decided it was time to combine?

extracts luminance in that situation, but I thought it would grab sum from all of them).

That's what it is supposed to do and what its purpose is, but squashing and stretching the dynamic range, detail and the per-pixel SNR prior, doesn't make a whole lot of sense.

I can't really tell much of a difference where you used Ha as lum there from some of the other versions though honestly.

Then I am frankly not sure what you are after any more... It shows all the detail you were after that you indicated in this post and is better(!) visible than in your PI/ST hybrid on all my screens. I am really wondering if there is something amiss with your screen calibration? Do you have, say, a phone or tablet handy? What do the images look like on those?

For RGB those times make sense as being important since each contributes directly to a proportional section of total luminance, but in narrowband Ha is always going to be hugely dominant and I don't think everyone always wants that in their final image.

Indeed, hence the recommendation to base your exposure times based on which detail you wish to acquire. However, given the renditions you prefer, it appears you actually do want Ha to be dominant in your image (indeed, it tends to be the most prevalent in terms of detail).

Ha is almost always going to be the dominant signal, and generally needs much less exposure time to achieve similar SNR to oIII and SII

Signal you acquire, constitutes simple succesful photon counts (not every count is successful, which is the origin of the shot noise).
Adding up the photon counts in the different bands just gives you more photon counts. It does not add noise depending on the band (unless a fixed amount of noise is present - thermal or shot noise from skyflow, etc.). This is the main reason for why you would make a linear synthetic luminance frame, weighted according to exposure times . And it's also the reason why stretching channels individually and then mixing them into a luminance frame yields wildly varying noise signatures.

Photosites are like electron "buckets". You have three buckets per pixel, per exposure. When creating a luminance master, then, for each exposure, you pour the three buckets into a one new bucket (adding the photon counts).

I have heard that as a technique to focus on ha and then shorter on the others and just using ha as lum, but I've been disappointed in the past not gathering enough signal on the other channels,

Not enough in terms of detail? Color? Were the objects poor in Ha signal/detail?

I don't really want to just use Ha as lum, and I think there is great detail in Sii I would want to bring out.

That is absolutely true for this object. In fact, as a reference, any object you see around on the Internet that was rendered in SHO (and was properly processed!) will show great detail in S-II, if you can see that nice orange/golden coloring in it (e.g. the mixture of Ha and S-II; red and green). In a properly processed image, the colors mean something.