StarTools

Hi everybody,

despite the fog season here I got almost two nights for imaging the Heart Nebula over the course of a month...


This is about three hours of broadband data enhanced with ten hours of duo NB data by using the NBaccent module. I really start to like this workflow for nebula images. Initially I thought this might be only (but very!) useful for emphasizing HII regions in galaxies. But it's also great for pushing details in nebulas without losing the natural look of a broadband image.

Here's another rendition, using the exact same data but with the duo NB for luminance and chrominance of the basic image and just adding broadband stars:


Also has cool stuff in it but I tend to prefer the NBaccent version.

Any other experiences with NBaccent on nebula images? Would be interested to see more of them and how different people use the module. I am also wondering if the ratio of the data is wrong for using it in NBaccent. Maybe I should have captured more broadband in comparison to NB since broadband gets processed with contrast, HDR, sharpen, decon, etc., while NB data is just wiped and stretched... So would it have been better to capture ten hours broadband and three hours of NB instead of the other way round like I have done. Any thoughts or experiences?

Regards
Stefan

PS. See astrobin for technical details: https://www.astrobin.com/2pahbr/0/

That visual spectrum one in particular is a stunner! Looks like very effective use of the NBAccent module. Well done!

The broadband version with NBAccents is superb!
If you have any working tips on the module use for large nebula, please pass them on!

Thanks for your kind words, Ivo and Andy!

Andy, I don't have any tips since I haven't used NBaccent a lot (apart from the Flaming Star Nebula viewtopic.php?f=11&t=2374) and haven't done dedicated comparisons So I don't know any best practices, but I can tell you how I've done it.

For NBaccent compositing I haven't chosen a ROI and threshold was set to 0 so most of the NB data was used for boosting. Afterwards I used the [HII/Balmer Series (Red/Purple) + Hb/O-III (Cyan)] option for response simulation and set the strength to 120%. I played a lot with the different response simulations. Red/purple looks much more appealing to me than the pure red while I couldn't see a major difference between teal and cyan for OIII. I also fiddled with the band balance... the preset 50/50 appeared best to me.

I didn't stick to the recommended workflow since I used NBaccent and AFTERWARDS went into SuperStructure instead of the other way round. My reasoning was that the NB data would introduce details which should be processed by SuperStructure so SuperStructure was the final module before noise reduction. As already noted I haven't done a specific comparison on which order is best, but this appeared reasonable to me and I was happy with the result.

I reproduced the processing based on the log and here is the image during particular steps of processing.

Before NBaccent: After NBaccent: After SuperStructure (DimSmall and Saturate): Afterwards I only did a noise reduction for the final image.

Here's the log for NBaccent and SuperStructure:
Hope this gives an impression of the workflow.

In my eyes the difference between the before/after NBaccent is huge and NBaccent makes a big improvement. It's ridiculously easy and very versatile. A real dream would be if one could also process the NB data in terms of contrast, HDR, sharpening, decon, etc. But I suppose that would be a major development project...

I still wonder what is the "correct" balance of exposure time of the broadband and NB data sets. Does it make sense that NB data has three times more exposure time than broadband like in this case? Or should it be the other way round? Or the same? Ivo, can you make a statement on this based on the algorithms used (@admin)? Or is there no right or wrong, but rather an "it depends"? But if so, on what does it depend

Best regards
Stefan

Thanks Stefan, thats very helpful information.
I agree on your thoughts of superstructure after the NBAccents, makes great sense to me!
Thanks for taking the time to detail your process.

Nice images!

I don't have much experience with NB Accent yet, having only really used it on one target, and on that one it was also a full boosting of both L and color, with some adjustments.

We probably need Ivo to clarify, but my understanding from earlier on is that the NB set will be double secret parallel processed in the background, same as color apparently is even though we are working on the L (assuming compose mode, of course). So those modules - Contrast, HDR, etc., will be taken into account and tracked, even though we get a bite at a separate AutoDev like restretching of the NB when we finally get to that module. No?

That seems like it would be the reasonable way to go about things too. The only module I'd like to see have separate effects is Wipe, since, were I to separately process my NB and RGB datasets, the Wipe settings would be quite different. For example RGB often requires a pretty good hammer on the gradient aggressiveness, but I wouldn't want to do that to the NB data, which may very well be amenable to using the NB preset and letting much of it float through.

Of course I may be completely misunderstanding what is happening... But since you can tri-toggle the color, L, and NB view in Wipe, that seems to be the intent?

Hi Mike,

Mike in Rancho wrote: ↑Sun Nov 14, 2021 6:10 pm We probably need Ivo to clarify, but my understanding from earlier on is that the NB set will be double secret parallel processed in the background, same as color apparently is even though we are working on the L (assuming compose mode, of course). So those modules - Contrast, HDR, etc., will be taken into account and tracked, even though we get a bite at a separate AutoDev like restretching of the NB when we finally get to that module. No?

My understanding is that the answer is "no" The documentation says the following (https://aife.me/startools18/modules/narrowband-accents):

The narrowband dataset is processed in parallel during your workflow; the Bin, Crop, Mirror, Rotate and - most notably - Wipe modules all operate on the narrowband accent dataset in parallel as you process the main luminance (and optionally chrominance) signal.

I think that HDR etc. would be mentioned here if they were also be applied to the NBaccent data.

I guess that I want to use the module in a way it is not developed for, namely for more luminance contribution. From the documentation again:

If you wish to use the narrowband signal as luminance or chrominance itself, rather than for accentuating luminance or chrominance, then the NBAccent module will not apply, and you should use the Compose module to load your narrowband as luminance and/or chrominance instead.

Unfortunately I am not able to get the star colors I want if I use the l-enhance data for luminance and broadband for RGB. Somehow I don't get the same colors as when I am using broadband for luminance and chrominance. I understand that this has something to do with LRGB emulation and how colors are perceived etc. Unsatisfying still... Star colors are much better with NBaccent while I still get luminance contribution from NB data. So I like this a lot. But I'd like to further process the NB data in HDR etc. too. Maybe I will try to go to color and NBaccent immediately after the global stretch and THEN go through my usual workflow with contrast, HDR, sharpen, decon, shrink, SuperStructure and denoise. Thus the NB data should get processed, too. Not exactly what NBaccent is intended for, but I might give it a try..

Mike in Rancho wrote: ↑Sun Nov 14, 2021 6:10 pm The only module I'd like to see have separate effects is Wipe, since, were I to separately process my NB and RGB datasets, the Wipe settings would be quite different. For example RGB often requires a pretty good hammer on the gradient aggressiveness, but I wouldn't want to do that to the NB data, which may very well be amenable to using the NB preset and letting much of it float through.

I agree on that one!

Regards
Stefan

Better documentation reading than I did. You may be right here, Stefan.

Perhaps I read more into "parallel processing" than was warranted. Though it seems to me that bin, crop, and rotate would fall into the "goes without saying" category.

Maybe I'll test it by doing something blatantly obvious in one of the substantive processing modules, and see if that carries through into the NB dataset when looked at during it's independent AutoDev.

Just in case anybody is interested: I applied a very similar workflow to the data than before but went to color and NBaccent immediately after the final stretch. Then I visited Contrast, HDR, Sharp, SVDecon, Shrink and SuperStructure in order to process the NBdata in the different modules. Afterwards Denoise. Here are the two images:

Recommended workflow with NBaccent rather at the end of the workflow:
NBaccent very early in the workflow: The most obvious difference is that the new rendition is not as soft as the first one. I guess the reason is that in the new one the contrast of the NB data is processed in the Contrast module while in the first one it is not. Apart from that in the new rendition aren't more details to see than in the second one like I had hoped. You probably can't tell since the images are too compressed but I wasn't able to see major differences (apart from global contrast) in the TIFFs, too. Maybe a tiny bit more detail in the central structure...not really sure.

So processing the NBaccent stuff with HDR, Decon etc. didn't result in substantially more detail but in a less soft image. This is of course a matter of taste. I prefer the first rendition which is based on the recommended workflow.

Regards
Stefan

Hi Stefan, nice testing.

Colors are a little deeper in the alternate workflow also. That said, it is often difficult to "identically" and comparatively process in ST with differing workflows, since the way it acts in pixels is dependent upon other modules actions, even with tracking. I think Ivo explained that recently regarding deconvolution?

I tried some quick testing too, and seems you are right about the documentation. I applied a strong shadows HDR to the data, which did not seem to carry over into the NBA. So perhaps the parallel processing is really just the housekeeping tasks like bin and crop. I did not test SVD, though that seems it might be more critical to match up, at least pertaining to stars.

Anyway, in my test data (RGB and Bicolor of Pacman) I noticed that applying 100% L contribution to NBA washed out all the details that were brought out in enhancement. Likewise, 100% color contribution "messes" with the way the RGB data looks. It seems somewhat akin to just composing the bicolor set as the L instead of NBA - the differences being that as L, it can be enhanced with things like contrast and HDR, but as NBA, you can adjust stretch, strengths, and contribution levels. Your alt-method is probably somewhat like the former.

I suppose Ivo would tell us - do you have a plan for combining RGB and NB, and does it make sense?

Using NBA in this way is probably a little bit outside of its design intentions.