40 hour Heart Nebula in Hubble HSO

[tooth_dr 7,826]

I scrapped all the Oiii and Sii data I previously took during a full moon (about 15 hours worth) and retook it all when the moon was a bit smaller at 76%.  Ha was taken during 98% and 67% moon.  All the lights were taken on the following nights: 12th, 19th and 20th September 2019.

Integration times, all in 600s subs unbinned:

Ha = 28.33 hours

Oiii= = 5.67 hours

Sii = 5.67 hours

 

The Ha data is really nice, and unsurprisingly the Oiii and Sii is not as strong (or nice).

I'm missing that (vital) step in my processing routine of getting the Sii and Oiii properly stretched to match the Ha, before combining.  I dont really know how to deal with the weaker data properly.  Any pointers would be appreciated.

What I do currently:

All the data is loaded into APP into separate channels/sessions.

The data is stacked and registered against the best Ha sub

This produces individual stacks of Ha, Sii and Oiii that are all registered

Each channel is processed with DPP in APP and then saved as a 16bit TIFF

Each is opened in PS

Stars removed with AA and any remnants removed and tidied up

I then open a blank RGB document in PS

I paste Ha into Green, Sii into Red and Oiii into Blue

Adjust the selective colour settings to get 'Hubble palette'

Adjust levels, curves, saturation until looks ok

All the Ha Sii Oiii data is then combined together in a single 'super' stack in APP using quality weighted algorithm to create a 'luminance'

That luminance layer is adjusted using levels, curves, and NC tools such as local contrast enhancement and deep space noise reduction (using masks to apply as required)

The luminance is pasted onto the above colour layer, and incrementally added using gaussian blur

Cropped and saved.

 

 

Here it is anyway   I haven't intended on any more exposure time for this one, but will consider it, if the expert opinion dictates otherwise!

 

CS

Adam

 

Edited September 22, 2019 by tooth_dr

[Craney 2,384]

Epic !!!!!!   

Just look at all that stratified structure  and  depth associated with Melotte 15 .     

I need a cold flannel on my head just thinking about the work you have put into this.   Great post.

[teoria_del_big_bang 1,584]

I know you are not happy but that looks a very nice image.

I was on the same target Friday and my Ha I was very happy with considering the moon and neighbours floodlights on and off all night but I don't really have anything much on my OIII or SII (from what I have seen on individual stretched subs - not had time to process them yet) so will also be interested on replies to your thread. Mine were all 400 S subs.

Steve

[knobby 3,646]

Really nice !

[vlaiv 5,695]

Very nice image indeed.

I have a few remarks, but don't take them as criticism of your work - just general observations.

Although I like the achieved palette, I would not call it SHO. This is something that is generally applicable to "SHO" renditions amateurs do, but your work here is a good showcase for my point, so I'd like to use it for that.

Most of us start by assigning SHO to RGB channels and then proceed to tweak things to be visually satisfactory (often influenced by previous work of others and online tutorials - there is now a "strong" sense of how should the image in "SHO" palette look like), but we end up in loosing what the SHO palette is all about - identifying different gaseous regions of the object - notably presence of hydrogen, sulfur and oxygen gas.

Here is what I mean - SHO maps SII region to red channel of the image, hence one would expect if there is distinctly red part of the image - that it should contain mostly SII gas. Or in another words - SII stack should have signal present in that place while Ha and OIII should lack it. Since you kindly provided stretched versions of all stacks - let's look if it is the case.

Top right corner shows structure that is particularly red:

But look at SII stack in the same region:

I'm not seeing that signal in there. Nor in OIII for that matter:

If we look at Ha stack on the other hand - above structure, rendered in red is clearly present and very strong:

What I believe is happening in this particular case is the fact that background level of SII is higher than other two after stretch and Ha luminance is strong, so mixing those two makes red color in those regions - not because SII signal, but because of SII background.

You can clearly see this in following region:

This is lower central part of Ha - background is very nice and black here - no signal captured. Same regions in OIII and SII:

You can almost make out same structure, but background is fairly grey here instead of being black as in above segment.

There is room for improvement here if you follow my argument and want to add that level of information to your rendering - try to do similar level of stretch (actually Ha is often much stronger as you've seen from your capture and is stretched quite a bit less than other two channels) - particularly pay attention to background levels being equal.

1 hour ago, tooth_dr said:

Each channel is processed with DPP in APP and then saved as a 16bit TIFF

I advise strongly against using 16bit per channel in any segment of the workflow - use 32 bit if you can for all intermediate steps.

1 hour ago, tooth_dr said:

All the Ha Sii Oiii data is then combined together in a single 'super' stack in APP using quality weighted algorithm to create a 'luminance'

I would advise against this. Use of synthetic luminance is fair processing technique, but one needs to be careful of how to create good synthetic luminance in narrow band data. One, most frequently used way is to just use Ha data. Ideally you want to make it by adding all three stacks while still linear - that will provide most accurate luminance. Problem with this approach is that Ha is high SNR data while other two are low SNR data - by virtue of having much lower signal to be captured in the first place. Adding stacks together does indeed add signal together but it also adds noise and you end up with result that has lower SNR value than single Ha channel.

Weighted algorithm does produce slightly better result, but in all likelihood - you will still end up with lower SNR then using Ha stack alone as synthetic luminance.

Good approach for creating synthetic luminance is - look at stretched versions of all three stacks, and observe if there is OIII / SII signal that is independent of Ha signal - it will rarely be the case, in most cases there will be Ha signal (that is stronger) in all the places there is SII or OIII. If Ha covers it all - just use Ha as luminance.

If there are regions with SII or OIII where there is no Ha (often in SN remnants like Crab where gasses move at different speeds due to different mass) - use "max" stack of three stacks. It should provide better SNR over addition. Alternatively you can only use "max" on region where there is SII and/or OIII and Ha for the rest (combine using masks).

[Xiga 1,698]

Looks amazing Adam. Very rich in detail and colour, top stuff

Are you doing gradient reduction on all 3 stacks prior to combining? It's crucial that you do. Vlaiv is also bang on about the background levels. After stretching all 3 channels, just pick an area of no nebulosity and if needs be bring in the black point to make sure the background level of that area is the same in all 3. 

Seeing as you already have Annie's Actions in PS, you should just use the 'Hubble Palette Creation' action to combine the channels. It does a lot more than just assign the channels to RGB.

Finally, I strongly recommend looking into Starnett++ for star removal. It does take about 20 mins per channel, but it's fully automated and the results are FAR superior to the AA action, and with hardly any cleanup needed afterwards. 

Congrats on an already cracking image! 

Edited September 23, 2019 by Xiga
Spelling

[tooth_dr 7,826]

Thanks for the comments everyone.  Thanks @vlaiv for that comprehenive overview

 

Just now, Xiga said:

Looks amazing Adam. Very rich in detail and colour, top stuff

Are you doing gradient reduction on all 3 stacks prior to combining? It's crucial that you do. Vlaiv is also bang on about the background levels. After stretching all 3 channels, just pick an area of no nebulosity and if needs be bring in the black point to make sure the background level of that area is the same in all 3. 

Seeing as you already have Annie's Actions in PS, you should just use the 'Hubble Palette Creation' action to combine the channels. It does a lot more than just assign the channels to RGB.

Finally, I strongly recommend looking into Starnett++ for star removal. It does take about 20 mins per channel, but the results are FAR superior to the AA action, and with hardly any cleanup needed afterwards. 

Congrats on an already cracking image! 

Cheers Ciaran.  I'm going to go back to the start with this one, it looks way over worked now that I've looked at it again.

I spent time doing light pollution removal tool in APT, to remove any gradients. Is that sufficient?  Plus I never thought of using the hubble palette creation tool - it's on NCs tools too i think d'uh.

Starnett++ isnt one I've heard of.  I will get that on the case later.

 

I think there is a lot more to get from this image, I hope I can get there !

 

 

Regards

Adam.

 

 

[Xiga 1,698]

4 minutes ago, tooth_dr said:

Thanks for the comments everyone.  Thanks @vlaiv for that comprehenive overview

 

Cheers Ciaran.  I'm going to go back to the start with this one, it looks way over worked now that I've looked at it again.

I spent time doing light pollution removal tool in APT, to remove any gradients. Is that sufficient?  Plus I never thought of using the hubble palette creation tool - it's on NCs tools too i think d'uh.

Starnett++ isnt one I've heard of.  I will get that on the case later.

 

I think there is a lot more to get from this image, I hope I can get there !

 

 

Regards

Adam.

 

 

Yeah the Light Pollution Removal tool in APP is the one to use on each, so you're good there. 

This is a Deep data set. Could be worth posting the raw data and see what some of the SGL gurus can come up with. Then you could zero in on the specifics of what they did. 

[tooth_dr 7,826]

12 minutes ago, Xiga said:

Yeah the Light Pollution Removal tool in APP is the one to use on each, so you're good there. 

This is a Deep data set. Could be worth posting the raw data and see what some of the SGL gurus can come up with. Then you could zero in on the specifics of what they did. 

I’ll upload the data later. I’m downloading Starnet now.  Cheers.

[alan potts 5,532]

Really nice image Adam so detailed. Do you ever wonder after 40 hours exposure what 60 hours looks like, asks he who did 2 hrs 30 mins last night and thought that was a lot.

Alan

[tooth_dr 7,826]

14 minutes ago, alan potts said:

Really nice image Adam so detailed. Do you ever wonder after 40 hours exposure what 60 hours looks like, asks he who did 2 hrs 30 mins last night and thought that was a lot.

Alan

Thanks Alan.  I would contemplate hitting another night of Oiii and Sii, perhaps in 900s subs.  Bear in mind, on a good clear night, I can get a total of 16+ hours.

[Skipper Billy 4,219]

Outstanding - and great patience !!!

I am getting better at gathering data but processing is still my weak point so I have nothing to add except it might be worth posting up your three stacked files to see what the Pixinsight and PS wizards can get from it then ask the questions about what they did they you didnt ??

May I ask why you gathered so much more Ha than Sii and Oii  ??

 

[MarsG76 4,791]

Excellent work...

[Jkulin 1,601]

Very nice Adam, nothing wrong with that.

[Lucas_M 262]

Wonderful image!

Congrats!!