Rosette Nebula: a first attempt at "Funny Colour" (HaHaRGB)

[Filroden]

This is my first attempt at taking and using narrowband within my LRGB images having recently acquired a Ha filter to fill the last slot in my 5 slot ZWO EFW.

I had taken a very little LRGB data already in late December and after an initial mistake in extracting the background (i.e. removing 80% of the nebula too) I ended up with the following:

(15s subs at -20C, 139 gain and 21 offset: 77 x L, 32 x R, 39 x G, 40 x B on the Skywatcher Esprit 80 with the ZWO ASI 1600MM-C and ZWO LRGB filters)

On Tuesday (23 Jan) I had my first opportunity to use my new Ha filter and chose the Rosette Nebula as my target. My intention was to capture Ha and much more LRGB. However, I have a limited imaging window as I can only observe between NNW and SSE, so the Rosette quickly passes through it. Another issue I'd also not taken account of was my neighbour's heating vent sends steam in roughly the SE direction, further limiting my window until it turns off.

So, I only achieved a little data to add to the above. All new images taken at -20C, 300 gain and 50 offset. During processing I discarded all L data as it was weak in comparison to the Ha data, even when blended. I have also blended the Ha data into the red channel using the tutorial on the Light Vortex Astronomy website. 

Ha: integration of 35 x 30s (new) and 22 x 60s (new)

R: integration of 33 x 15s (original) and 10 x 30s (new)

G: integration of 39 x 15s (original) and 10 x 30s (new)

B: integration of 30 x 15s (original) and 10 x 30s (new)

Image captured in SGPro and processed in Pixinsight and finished in Lightroom.

 

Rosette in Ha

 

Rosette in Ha-R(Ha)GB

 

I know there is far too little data but I'd like advice on combining Ha data with LRGB. I've used the tutorial on Light Vortex Astronomy which takes (I think) a weighted maximum of the data between the two mono images being processed (L & Ha or R & Ha). However, when combining Ha to the L I found it washed out the more detailed Ha data. Likewise, when combining with the R channel, I found that star colours seemed to be okay, but the nebula remained mono-chromatic only now in red, not white.

So my resulting image uses the Ha as luminance and the Ha data also dominates the red channel. As these two data sets match, they've effectively reduced me to a red tinted monochrome image, though with some nice small stars which retain some RGB colour. Are there better ways to combine Ha data or is is it just I don't have enough RGB in comparison to Ha?

(Oh, and I know how badly spaced/tilted my image is - still trying to figure out how to solve that one)

Would love advice/examples of how to start on incorporating Ha into my workflow.

[Ibbo!]

nice result, you seem tpo have overcome a lot of hurdles to get yoiur data.

[AKB]

You probably can't do better than follow Olly's advice here:

 

[Filroden]

8 minutes ago, Ibbo! said:

nice result, you seem tpo have overcome a lot of hurdles to get yoiur data.

Thank's Steve. Indeed, I don't think this hobby would be as satisfying if it wasn't for all the difficulties

8 minutes ago, AKB said:

You probably can't do better than follow Olly's advice here

Ah, that makes more sense. One of Photoshop's key advantages over Pixinsight is that ability to blend "on the fly" seeing the results in real time, rather than trying to work it out through equations. Once I've had chance to capture more RGB I will give this a try.

[alan4908]

Ken

That is an impressive image for a first attempt at incorporating Ha data !

1 hour ago, Filroden said:

Would love advice/examples of how to start on incorporating Ha into my workflow.

You have above some excellent suggestions in the above (repeated) post by Olly. Some additional suggestions that you might wish to consider:

1. Experiment with PS screen as opposed to PS lighten as the way to blend the Ha data. I've found that sometimes one works better that the other, this depends on the data set.

2. Sometimes I notice that the star colours are impacted by the Ha data, to correct for this you can take a very lightly streached RGB layer (so that you can see the star colours). Slightly black clip this and then PS screen this with your LRGB result. 

3. Try adding a bit of green or blue after you have the LRGB result to change the type of red that you like most. One way to do this is via the Hue/Saturation command in PS, just move the slider (slightly) to either the blue or green direction and adjust to taste.

Alan 

[gnomus]

I think it is a good image of the Rosette.  I don't really know enough about these new CMOS cameras so cannot comment about exposure times and total integration time.  I assume that 'the more data the better' rule still applies.  

One of the things I picked up on in Warren Keller's PI book is that perhaps you don't need to stretch the L, RGB and Ha stacks to the max before combining them.  You might find that less stretching would retain some of your colour.

For incorporating Ha into LRGB, I use the method shown to me by Mr Penrice.  I do this in Photoshop, I don't know if you can do it in Lightroom.  

1. Bring up the Channels Tab
2. Split Channels
3. Copy and Paste the Ha on top of the Red Channel
4. Change Blend Mode to Lighten
5. Flatten
6. Merge Channels (RGB)  

After this you can add Ha as a Luminance layer, but dial the opacity back to 10-20% (Or don't add Ha as Lum at all).

Sometimes this will turn the reds a salmon pink.  If it does, you can do Selective Colour-Red and dial back on the Cyans to taste.

I hope that is helpful.

Steve

[michael.h.f.wilkinson]

Very impressive given all the difficulties

[ollypenrice]

11 minutes ago, gnomus said:

I think it is a good image of the Rosette.  I don't really know enough about these new CMOS cameras so cannot comment about exposure times and total integration time.  I assume that 'the more data the better' rule still applies.  

One of the things I picked up on in Warren Keller's PI book is that perhaps you don't need to stretch the L, RGB and Ha stacks to the max before combining them.  You might find that less stretching would retain some of your colour.

For incorporating Ha into LRGB, I use the method shown to me by Mr Penrice.  I do this in Photoshop, I don't know if you can do it in Lightroom.  

1. Bring up the Channels Tab
2. Split Channels
3. Copy and Paste the Ha on top of the Red Channel
4. Change Blend Mode to Lighten
5. Flatten
6. Merge Channels (RGB)  

After this you can add Ha as a Luminance layer, but dial the opacity back to 10-20% (Or don't add Ha as Lum at all).

Sometimes this will turn the reds a salmon pink.  If it does, you can do Selective Colour-Red and dial back on the Cyans to taste.

I hope that is helpful.

Steve

Just to add that, at stage 4, you can give the Ha a further stretch if it looks as if it isn't giving as much as you'd like. Just pin down the background in Curves, though, so as not to lift the background sky values in red. 

Olly

[Filroden]

So if I have this right, blend lighten mode allows the brightest pixels through? So if I wanted to achieve the same in PI I need to combine the two channels using the Max operator?

@alan4908 also suggest trying Screen. How does that work?

[Filroden]

25 minutes ago, gnomus said:

I assume that 'the more data the better' rule still applies

Always  But the low read noise and good sensitivity on the ZWO ASI1600 does allow a passable result from limited data so even limited imaging time means I can do something. Certainly it's more forgiving than a lack of data using a DSLR.

[gnomus]

9 minutes ago, Filroden said:

So if I have this right, blend lighten mode allows the brightest pixels through? So if I wanted to achieve the same in PI I need to combine the two channels using the Max operator?

@alan4908 also suggest trying Screen. How does that work?

Yes the 'brighter' of the two corresponding pixels in the two layers/images is what is used.  I cannot comment on the PI method.

[Filroden]

2 minutes ago, gnomus said:

Yes the 'brighter' of the two corresponding pixels in the two layers/images is what is used.  I cannot comment on the PI method.

Cool, then I think I can achieve the same effect in Pixinsight. I try to avoid swapping programmes mid workflow (the exception being to improve masks) as I have to convert file formats and I'm never sure if that introduces problems (not least that PI can work in 32bit for every process but PS only seems to work on 16bit for some processes). I like to do as much early processing in PI then move into PS or even directly into Lightroom for final touches.

[gnomus]

If you work out an equivalent method in PI, I'd be happy to hear about it, for the reasons you state.  There is a PI script that allows you to add NB to RGB too.  I have only had a little play with this.

I do find myself jumping back and forth between different programs all the time.  I know what you mean about 32 bit vs 16 bit, but 16 bit is still pretty deep and the images we post are 8 bit.  I don't know if anything significant is 'lost' by going to 16 bit earlier in the workflow.  It would be inresting to hear what others think.

[alan4908]

1 hour ago, Filroden said:

So if I have this right, blend lighten mode allows the brightest pixels through? So if I wanted to achieve the same in PI I need to combine the two channels using the Max operator?

@alan4908 also suggest trying Screen. How does that work?

Ken

The Photoshop Screen blending mode is an additive function and always results in a brighter pixel. Mathematically, if you have two layers A and B then PS Screen is equivalent to 1 - (1-A)*(1-B) so it is something you can easily code in Pixel Math within Pixinsight. 

You have to be particularly careful when using the PS screening function when using Ha with Lum,  this is because you might make things too bright and hence you would destroy colour.  To explain: if you have a total Luminance above about 200 in PS then it will be difficult to colour, so you need to be careful not to add either too much Ha or conversely make the Lum too bright before you blend the Ha. 

Alan

[Filroden]

41 minutes ago, alan4908 said:

You have to be particularly careful when using the PS screening function when using Ha with Lum,  this is because you might make things too bright and hence you would destroy colour.

But I could normalise the data back to a scale from 0 to 1 and then review contrast from there to avoid blowing out the colour?

[alan4908]

1 hour ago, Filroden said:

But I could normalise the data back to a scale from 0 to 1 and then review contrast from there to avoid blowing out the colour?

Photoshop displays pixel values in 8 bit so 0 to 255 - if we say don't beyond approx 200 in total Lum to avoid destroying colour whilst in PS, we are effectively saying don't go beyond 200/255 = 0.78 whilst in Pixinsight, which by default displays pixel values between 0 and 1.  So, all you need to do is to ensure that when you perform the Pixinsight Screen function via Pixel math (eg add the Ha to the Lum) that you don't go more than this value with your total Lum (eg L + Ha) if you want colour in your image at this point.

Alan

 

[Filroden]

30 minutes ago, alan4908 said:

Photoshop displays pixel values in 8 bit so 0 to 255 - if we say don't beyond approx 200 in total Lum to avoid destroying colour whilst in PS, we are effectively saying don't go beyond 200/255 = 0.78 whilst in Pixinsight, which by default displays pixel values between 0 and 1.  So, all you need to do is to ensure that when you perform the Pixinsight Screen function via Pixel math (eg add the Ha to the Lum) that you don't go more than this value with your total Lum (eg L + Ha) if you want colour in your image at this point.

Alan

 

That makes sense. Thank you.

[PatrickGilliland]

Hi Ken - Steve and Olly beat me to many of my tips!  So I can take the next couple of minutes off........

So, HLRGB, try the HA in the lum channel it works better from my experience.  When you add yu are looking for the tipping point where red start to turn salmon Vs detail.  When you get there you can often complete a few iterations of colour adjustments and a little more ha, repeat gently a few times.

Another nice feature of using HA onto a LRGB is the fact you can with some careful masking add the stars as a lightness layer from Ha to LRGB, thus providing slightly tighter stars. Some careful masking and feathering will make that fairly easy.

Super first go though, certainly nothing wrong hopefully with all the tips here next version will please you even more.

Paddy 

[Filroden]

9 hours ago, PatrickGilliland said:

Hi Ken - Steve and Olly beat me to many of my tips!  So I can take the next couple of minutes off........

So, HLRGB, try the HA in the lum channel it works better from my experience.  When you add yu are looking for the tipping point where red start to turn salmon Vs detail.  When you get there you can often complete a few iterations of colour adjustments and a little more ha, repeat gently a few times.

Another nice feature of using HA onto a LRGB is the fact you can with some careful masking add the stars as a lightness layer from Ha to LRGB, thus providing slightly tighter stars. Some careful masking and feathering will make that fairly easy.

Super first go though, certainly nothing wrong hopefully with all the tips here next version will please you even more.

Paddy 

Thank you for the additional pointers  Now if only the weather would be so kind to allow me to collect more data so I can test out all the good advice.

@gnomus, I think I've got my PixelMath expression to mimic Photoshop's Lighten blend, including the opacity slider. Just replace the RefImage and TarImage with the correct image names. I tested this a couple of times and it seems to work the same. Opacity scales from 1 to 0 on the "top" TarImage layer.

[gnomus]

Thanks very much for posting this, Ken.  I'll certainly give it a try.  In PS 'Lighten' I generally leave the Opacity at 100%, so we may not need that feature, but it is good to know how to do it.  One advantage of doing this in PS is if you find that the Ha is having little or nor impact (or perhaps it is lightening the background too much), then, as Olly says, you can tweak Levels and Curves on just the Ha layer to get the result you want.  It would be great if PI could offer more 'Preview' functionality than is currently the case - I am using it more and more, but I still feel the need for the PS comfort blanket on most images.  

Steve   

[Filroden]

That's what I like in PS too. But it just takes a few more steps in PI. Apply the PixelMath, see if it works, adjust the Ha image with CurvesTransformation and reapply PixelMath. As PixelMath, in my example, is set to create a new image you can just keep repeating this and compare the various versions. 

That said, I'd probably create a copy of the Ha master and use that so I could always revert back to the original. 

[Barry-Wilson]

Ken

You may well know that there is a Blend script in PI with a menu choice of various blend modes using PixelMath to mimic the PS blend modes.  You don't see the formula behind the blends you just select the mode, eg Lighten, Soft Light, Hard Light etc and the percentage of blend between two images.

For adding Ha to RGB I will often use the MultChannelSynthesis HaRVB-AIP script perhaps setting the Ha image (Gray Mode) to 60% and Ha (RGB Mode) to 40% and then use PixelMath to blend the two resultant colour images.  Saturation is sometimes needed afterwards to achieve your desired result.  As an example, not meaning to highjack your thread, see my recent HaRGB Cone Nebula.

You have a very good HaRGB Rosette so you clearly have a good grasp of your processing goals and PI techniques.  Keep it up!

CS!

Barry

[Filroden]

16 minutes ago, Barry-Wilson said:

You may well know that there is a Blend script in PI with a menu choice of various blend modes using PixelMath to mimic the PS blend modes.  You don't see the formula behind the blends you just select the mode, eg Lighten, Soft Light, Hard Light etc and the percentage of blend between two images.

For adding Ha to RGB I will often use the MultChannelSynthesis HaRVB-AIP script perhaps setting the Ha image (Gray Mode) to 60% and Ha (RGB Mode) to 40% and then use PixelMath to blend the two resultant colour images.  Saturation is sometimes needed afterwards to achieve your desired result.  As an example, not meaning to highjack your thread, see my recent HaRGB Cone Nebula.

You have a very good HaRGB Rosette so you clearly have a good grasp of your processing goals and PI techniques.  Keep it up!

I didn't know that. And now I do, I can save myself a lot of time  Thank you!

I played with the HaRVB-AIP script (and the other two scripts in that same sub menu) but I couldn't really see what it was doing and the defaults didn't give me what I expected. I'll try your suggestion (I now have so many things to try if only I could get the clear night to gather much needed additional RGB data).

And I just took a quick peek at your site. Blimey! If I only achieved 30% of that quality I'd be over the moon. It was NGC1333 that inspired me to push harder with imaging. It's such a beautiful target and I'd seen so many great versions. I now count yours as one of them.

[Barry-Wilson]

More than happy to help with PI via PM Ken (as I have done with a number of members).

[Filroden]

Just had a quick play with the  HaRVB-AIP script using a saved (but poor) RGB image and the (better) Ha image. I have to say that it's done a much better job than using the PixelMath expressions I'd taken from the tutorial. I think that's largely because the tutorial assumes the quality of data is available, whereas I suspect the script is more forgiving!

It's not great (particularly in the fringes and background where poor colour casts have crept in due to the weakness of the original RGB data) but I like that this version has some colour graduation as well as lightness graduation within the nebula. If I follow this route for processing I'm going to have to deal with the halos but Patrick's given me a great way in PS to reduce/remove these.

I've also found the blend script (PJSR: Blend Script https://pixinsight.com/forum/index.php?topic=3902.0). Shame it doesn't offer a preview, but it will be a great time saver!