Adding Ha to OSC (or Lum?)

[900SL]

Hi

Just ordered a mono cam to run alongside an OSC

I'd like to try some Ha RGB or L RGB and would appreciate any pointers on how to do this in Siril.

I assume I need to somehow scale / register the separate stacks so they match ?

Siril offers  pixel math ( which i assume needs matched and scaled stacks) but also an RGB composition script, which I am unfamiliar with

Appreciate any advice on this

[900SL]

Found this so far from Oskari, which sets out how to calibrate and register different scales

 
https://stargazerslounge.com/topic/410888-combining-images-from-different-focal-lengths/?do=findComment&comment=4389104

[carastro]

I don't use Siril, but the main thing is to register the two images so they will exactly fit when combined.  I use Registar for this.   Generally Ha is added into the red channel.  

Edited December 20, 2023 by carastro

[ollypenrice]

I don't know how you'd do it in pixel maths (confound that American spelling!!) but, in Photoshop, I add it to the red channel in blend mode lighten. This means that it is only applied to the red channel where the pixel brightness is higher than the red. How might this be done mathematically? I really don't know but there will be a way.

A couple of pointers to Ha processing, though, when the result is destined to go into the red channel.

- You can ignore low level noise which is less bright than the red because it won't be applied.

- You can stretch the strong signal very hard indeed.

- Your background value in Ha should be a point below the red to avoid its being incorporated.

- It can be good to go for extremely high local contrasts because these will be diluted by the red channel.

In short, I wouldn't process an Ha image in quite the same way for adding to red as I would process it for standalone Ha.

Olly

[ONIKKINEN]

The pixel math route of adding Ha to RGB is really not very good, or easy to do for that matter. More ways you can make a mess than a result you like i would say. But if you have to, it happens by using the max (x,y) operator where you take the max value from 2 images combined to one result. Noise levels need to be similar and gradients need to be eliminated to have a chance at success, or you just add the noise and gradient from one image to the other (and the stacks need to be linear fit to each other).

If PixInsight is not available then Photoshop or Gimp is what i will recommend for this.

[ONIKKINEN]

Found a thread from a couple months ago where this Ha to RGB thing was discussed:

 

[Padraic M]

It might help your thought process to separate the original question into multiple questions:
1. how to register multiple channels from different cameras and scopes to a single reference frame and image scale
2. how to combine three OSC channels with one Ha channel to create a new three-channel RGB image which includes Ha to your desired recipe
3. as an advanced topic, and as detailed by @ollypenrice above, do you combine linear data only, or can you pre-stretch the channels to best advantage before combination?

 

For question 1:
I use Siril for stretching, but I don't use it for registration and integration. I use Astro Pixel Processor for that, and then import the linear integrated image into Siril. APP does all the hard bits relating to scaling and registering the images for you. It may be that Siril can do this too but I haven't tried it.

To build a HaRGB integrated image in APP, you need to first separate the OSC linear image into separate Red, Green and Blue channels. On the Calibrate tab, select 'Split channels'.

Then, integrate each channel separately to create Ha, R, G and B stacked images. You can register all lights to the same light frame at this stage if you like. Just make sure that the reference frame is integrated into its own channel, and then it is deselected but still marked as the reference frame so that it is not also integrated into each of the other stacks. (As an alternative, you can register and integrate them completely separately; then, reimport the integrated frames as Lights, register them together, save the registered frames and continue on as usual).

Note: When you are registering light frames from different cameras or scopes in APP you need to set some specific parameters on the Register tab, or the registering and integration won't work. The most important one is to deselect 'Same camera and optics'. It is also advised to select 'Use dynamic distortion correction'. If the two pixel scales and fields of view are reasonably similar this is probably enough. You can crop the channel images using the Batch Crop tool to remove any stacking artefacts. But if the scales and the FOV are very different, the techniques required become similar to creating a mosaic. There are good tutorials on the APP website if you need to get into this level of detail.
Also, having close to the same camera rotation angle on both setups is important, as even a small rotational difference can result in a lot of cropping around the edges.

 

For question 2:
There are many options for combining the four channels into a single image. In APP, after integrating each of the four channels, you can use the 'Combine RGB' tool on the Tools tab. You can play with different recipes or create custom ones here - try starting with HaRGB and playing with the sliders. Ha is usually added into the Red channel but you can experiment with how much of it you want to include, and also if you want to reduce the level of Red that comes from the OSC Red data. You can alternatively add the Ha data as a Lum channel; I find that adding it at much more than 20% washes out the OSC colour significantly. I do all of this is with linear data.

When you have saved the integrated linear HaRGB or LRGB image, you can close out of APP and continue processing (stretching) in Siril as you would do with any RGB image.

As the four channels are already registered, combining the channels can equally be done with Pixelmath in PI or Siril; or can be done with Photoshop or Gimp.

I don't think the RGB composition script in Siril will help - afaik that is for combining RG and B channels from a mono camera and filters into a single RGB image.

 

I'll leave question 3 to others with more experience!

[ollypenrice]

7 hours ago, Padraic M said:

 

I'll leave question 3 to others with more experience!

I don't want to present myself as a pundit but my view is that, if you combine it at the linear stage, you have relinquished almost all control over it.

Let's remember that there is no 'natural' value for the Ha contribution. There is no way to calibrate it 'scientifically' with the rest of the image because the calibrated Ha is, surely,  just what is already contained in the red channel.* The whole point of the addition of Ha is to distort its contribution to the image and allow us to reveal structures otherwise not visible, though certainly present in scene we are photographing. This is an artificial process and we are the the artists - or artificers - responsible.

Olly

*This does not apply to those daytime DSLRs whose filters do block most of the Ha.

[The Admiral]

When I was fiddling around in Siril to add Ha to a wide band image, I came across this filter function 'Contrast-Limited Adaptive Histogram Equalization (CLAHE)'. It really perked up the Ha channel before combining. Perhaps worth a look.

Ian 

[Rodd]

3 hours ago, ollypenrice said:

I don't want to present myself as a pundit but my view is that, if you combine it at the linear stage, you have relinquished almost all control over it.

Let's remember that there is no 'natural' value for the Ha contribution. There is no way to calibrate it 'scientifically' with the rest of the image because the calibrated Ha is, surely,  just what is already contained in the red channel.* The whole point of the addition of Ha is to distort its contribution to the image and allow us to reveal structures otherwise not visible, though certainly present in scene we are photographing. This is an artificial process and we are the the artists - or artificers - responsible.

Olly

*This does not apply to those daytime DSLRs whose filters do block most of the Ha.

Hmm. But if you compare a red stack with an Ha stack there are obvious differences. The red filter won’t pick up all Ha.  There is an overlap in frequencies, for sure, but they are not equal. Maybe I misinterpreted what you said?

[Padraic M]

4 hours ago, ollypenrice said:

I don't want to present myself as a pundit but my view is that, if you combine it at the linear stage, you have relinquished almost all control over it.

Let's remember that there is no 'natural' value for the Ha contribution. There is no way to calibrate it 'scientifically' with the rest of the image because the calibrated Ha is, surely,  just what is already contained in the red channel.* The whole point of the addition of Ha is to distort its contribution to the image and allow us to reveal structures otherwise not visible, though certainly present in scene we are photographing. This is an artificial process and we are the the artists - or artificers - responsible.

Olly

*This does not apply to those daytime DSLRs whose filters do block most of the Ha.

Absolutely Olly I agree, but you do still have some control over it, and processing at the linear stage is a relatively simple way to get started with HaRGB/LRGB/HaGB/(HaR)GB combinations etc. As you say, you are deciding how much you want to highlight the Ha content in your image at the expense of non-Ha red frequencies which is not a natural representation of the scene. You've given me food for thought on non-linear combination, work for another day!

@Rodd the Red stack contains all of the Ha light as well as other colours - the standard red filters on the market filter in a continuous range of wavelengths both shorter and longer than Ha at 656nm.

A narrowband Ha filter only allows through light at the Ha wavelength +/- half the bandwidth. It filters out all of the other red light that a standard red filter would have allowed through, but it doesn't include any light that is not transmitted by a standard red filter. So by selectively processing Ha and Red you are effectively emphasising that narrow band of Ha at 656nm by attenuating or even removing all of the other red wavelengths. All while (although not necessarily) leaving the green and blue colour levels untouched.

 

[900SL]

On 19/12/2023 at 12:20, carastro said:

I don't use Siril, but the main thing is to registar the two images so they will exactly fit when combined.  I use Registar for this.   Generally Ha is added into the red channel.  

Thanks C. You can register separate stacks and image scales AFAICT in Siril, using the manual processing tabs. I'm looking deeper into that now. 

[900SL]

On 19/12/2023 at 13:48, ollypenrice said:

I don't know how you'd do it in pixel maths (confound that American spelling!!) but, in Photoshop, I add it to the red channel in blend mode lighten. This means that it is only applied to the red channel where the pixel brightness is higher than the red. How might this be done mathematically? I really don't know but there will be a way.

A couple of pointers to Ha processing, though, when the result is destined to go into the red channel.

- You can ignore low level noise which is less bright than the red because it won't be applied.

- You can stretch the strong signal very hard indeed.

- Your background value in Ha should be a point below the red to avoid its being incorporated.

- It can be good to go for extremely high local contrasts because these will be diluted by the red channel.

In short, I wouldn't process an Ha image in quite the same way for adding to red as I would process it for standalone Ha.

Olly

Thanks Ollie. I went 'whoosh' when I read this:) 📚 some background reading required. 

[900SL]

On 19/12/2023 at 17:32, ONIKKINEN said:

The pixel math route of adding Ha to RGB is really not very good, or easy to do for that matter. More ways you can make a mess than a result you like i would say. But if you have to, it happens by using the max (x,y) operator where you take the max value from 2 images combined to one result. Noise levels need to be similar and gradients need to be eliminated to have a chance at success, or you just add the noise and gradient from one image to the other (and the stacks need to be linear fit to each other).

If PixInsight is not available then Photoshop or Gimp is what i will recommend for this.

I might make the leap to Pixinsight now I've got a mono camera. Thanks Oskari

[Elp]

Tbh, even though I use Siril for pre processing, something like this is so much easier to do in PS/Gimp or equivalent, you also have so much more control over the look as can make adjustments in real time as well as mask areas to only affect certain regions etc and also being non destructive (likely PI also allows this but at least the traditional software you can also apply the skills to any image or creative project, not just astro). See my latest heart and soul image as it's a similar process you're looking to do but I did it more artistic than realistic. Still took ages to do though.

Edited December 20, 2023 by Elp

[Rodd]

1 hour ago, Padraic M said:

Absolutely Olly I agree, but you do still have some control over it, and processing at the linear stage is a relatively simple way to get started with HaRGB/LRGB/HaGB/(HaR)GB combinations etc. As you say, you are deciding how much you want to highlight the Ha content in your image at the expense of non-Ha red frequencies which is not a natural representation of the scene. You've given me food for thought on non-linear combination, work for another day!

@Rodd the Red stack contains all of the Ha light as well as other colours - the standard red filters on the market filter in a continuous range of wavelengths both shorter and longer than Ha at 656nm.

A narrowband Ha filter only allows through light at the Ha wavelength +/- half the bandwidth. It filters out all of the other red light that a standard red filter would have allowed through, but it doesn't include any light that is not transmitted by a standard red filter. So by selectively processing Ha and Red you are effectively emphasising that narrow band of Ha at 656nm by attenuating or even removing all of the other red wavelengths. All while (although not necessarily) leaving the green and blue colour levels untouched.

 

Yes but there is a big difference between an Ha stack and a red stack. If you remove all but Ha from a red stack, it does not equal an Ha stack. For one thing, in my sky my red exposures are 1-2 minutes and my Ha exposures are 10-20 minutes.  Many structures that are made from ha frequencies do not have enough signal in that frequency in a red stack.  A good example is Ha regions in galaxies-like m106. One adds ha to galaxies because the red channel doesn’t have enough ha signal to do them justice. One would have to collect way to much red to see the same structures, and they wouldn’t be the same, anyway. You’d have to remove non ha signal to see them.  Otherwise, narrow we band filters would be worthless

[Rodd]

36 minutes ago, 900SL said:

I might make the leap to Pixinsight now I've got a mono camera. Thanks Oskari

Adding ha to the red channel is easy in PI using pixel math. You just have to know the method.  

[ollypenrice]

8 hours ago, Rodd said:

Hmm. But if you compare a red stack with an Ha stack there are obvious differences. The red filter won’t pick up all Ha.  There is an overlap in frequencies, for sure, but they are not equal. Maybe I misinterpreted what you said?

There's no reason why a red filter won't, in principle, pass all Ha but there is every reason why an Ha filter won't pass all red. The Ha will also exclude more LP in many cases. And then, this being the case, we can stretch the Ha harder as well.

Olly