How to assemble Ha and OIII DSLR on a RGB image

[Walcho]

Hi everybody,

I accumulated a few hours exposure on NGC6888 with my Canon EOS 1000D using different filters.
Distribution of exposure :
7:10 h in RGB
9:55 h in Ha
7:20 h in OIII

It's fine, but since I never had to do processing with having to assemble color layers, I have some questions.

How assemble the Ha layer and the OIII layer on my RGB image?

I did the preprocessing for each layer separately: RGB, Ha, OIII.
Stacking is in progress.

Once my Ha and OIII raw stacked, obtained from a color sensor, do I have to extract the color channel which interests me, red for ha and Blue for OIII?
Should I use the stacked raw like they are?
Should I extract the L-Channel from my Ha?

These are my questions. I hope to take advantage of your advice to avoid doing all these tests myself. 

Besides, another question comes to mind when writing these lines, should I extract the color channels at the time of preprocessing ?

And finally, to finish, what process or script is to use to combine my different layers? Ha-RGB? PixelMath? .... ?

Thank you in advance for your advice.

[Walcho]

Hello All !

Not easy image processing with filters!
It was much easier to use the raws coming from the DSLR, but in return, more signal in the red with the Ha filter and some new nebula appear with the OIII filter!


At least, I could see that the best method was to extract the color channel and use the red layer of the HA from the DSLR and that the green of the OIII-DSLR. Moreover, some processes or scripts work only with a monochrome image.

The problem is that it gives as many possible combinations to test.


Now I have tested so much that I do not know what I did, with good results, less well, strange colors etc ...
I also tested so many Process, scripts, formulas found on the net for PixelMath, that I do not know how I got certain results.
And I have tried linear and non-linear image processing and I can not say what is the best way to proceed.

Anyway, I will have to resume with methodology and set myself a process to follow that I can save, to notrepat all trial and error for a mediocre result.

An assembled image only in Ha-RGB.
An overview :

Then a zoom on the "crescent" ...

...and on the area where I hoped I could reveal the "Soap bubble".

And the image Ha-RGB-OIII very noisy and too saturated.
An overview :

Again, a zoom on the "crescent", where, magic OIII, we see appearing a bluish envelope ...

... and the area of the "soap bubble", where, if you tear-off your eyes you will see slightly the outline.

But there is still work to do, so tips, tricks and advice are welcome.

[Walcho]

Hello all,

I continue my post alone, nothing is sure but I have of big chance to be in agreement with myself.. 

So, I progress , I progress in my tests!
Following wise advice, I made a combination Ha + OIII-Green to do a Luminance.
OIII-Green rather than OIII-Blue, because more signal in the Green.

On the other hand my OIII layers being noisy, I will perhaps have to redo preprocessing on the OIII, because it rotten my Luminance a little.

Anyway, I did some integration tests L RGB Ha OIII, thanks to the SHO-AIP script, it gives interesting results. I slowly begin to understand how to do the processing. (hopefully).

I was even wondering if it would not be interesting to do a preprocessing on the RGB layers separately, just to clean the images even better?
that would mean re-start from the beginning ... but I'm not affraid anymore ...

[RickS]

Extracting the red channel from the DSLR/Ha to get a monochrome Ha master sounds like the right thing to do.  For Oiii, you could extract both green and blue channels and then use ImageIntegration to combine them (two copies of each file so you have more than three images, noise-weighted, no rejection.)  This should give you a little better SNR that just taking the green channel.

SHO-AIP is a good choice for combining everything but you could try NBRGBCombination as well.  There are many other options if you want to explore PixelMath, of course.

Cheers,
Rick.

[Walcho]

Hello Rick

Thank you very much for this answer.
I didn't think about combining Green and Blue together! This is a trail that seems really interesting and that I will explore.

But, I don't understand what you mean by:

two copies of each file so you have more than three images

I didn't know NBRGBC, but as I am curious, I will test it too.
I have already explored PixelMath, I love but I'm not an expert.
I just follow tutorials found on internet and try to understand to adapt them to my needs.
But I will persevere in these ways.

Thanks again.

[pfile]

oh it's just simply that ImageIntegration requires 3 input images at a minimum. so if you are trying to create a lum master from two channels, you need to add at least a 3rd file. if you added just one of your files twice, the weighting in the final image would be wrong, so you add 2 copies of each image.

rob

[Walcho]

OK, I think I understand what you are explaining.

This is based on the assumption that I extract the G and B channels from my final OIII-RGB image, is that right?

But my idea was to extract the G and B channels of each OIII-RGB raw image and preprocess them separately.
At the integration I will have more than 3 images.
Wouldn't it be more efficient ?

[pfile]

well i muddied up what rick was saying - he's describing how to combine the green and blue channels into a single OIII master. i used the example of an L master but that was not relevant to your issue. regardless you always need at least 3 input images to ImageIntegration. since in this case you have the G and B channels of the original master, you need to add each one twice so that they contribute equally to the final integration.

i think what is being suggested:

1) extract the red channel from your master integration, to serve as a Ha master
2) extract the green and blue channels from the master integration, then integrate them together to serve as an OIII master
3) create a bicolor RGB image from the two masters, perhaps R=Ha, G=OIII, B=0.2*Ha+0.8*OIII (some people do this since in theory wherever there is Ha signal, there is Hb signal in approximately the ratio given here...)

indeed if you are willing to split the raw files, you can extract the G and B channels and then integrate all of the G and B subs together to get the OIII master. if you really want to get technical, you can use SplitCFA to get the two unique G channels and integrate all the B, G1 and G2 subs together, but at that point you are going to be working with the equivalent of SuperPixel debayered data - half the dimensions of the standard debayered images.

rob

[RickS]

Thanks for chiming in, Rob.

But my idea was to extract the G and B channels of each OIII-RGB raw image and preprocess them separately.
At the integration I will have more than 3 images.
Wouldn't it be more efficient ?

You could try that but you'll need to do rejection and I suspect that differences between the G and B channels may cause issues.  Never hurts to experiment a bit, of course.

[pfile]

i don't think that would be a problem - i know from my DSLR days that the G and B images in an OSC thru an OIII filter (7nm) were essentially identical. of course YMMV.

but of course just integrating all the Gs and Bs separately and then integrating the two masters would work just as well.

rob

[Walcho]

Thank you Rob and Thank you Rick,

this is a specialized exchange for me and very interesting.
I just discovered the process SplitCFA thank you for the info, it will be much easier and more effective than extracting each layer of each image ... 
I am also pretty confident about the rejection between the blue and green channels because they're coming from the same RGB image, but maybe the reasoning is not correct.

I will do my tests and I will keep you informed.

[Walcho]

Hello Guys,

I would need again your precious advice.

What do you think is the best way to proceed ?

Split the CFA channels directly after Lights Calibration and before CosmeticCorrection ?
or
Split the CFA channels after CosmeticCorrection before Alignement ?

And one more question, when should I debayer my images ?

Thanks in advance !

[sharkmelley]

For what it's worth I would do it differently. Don't do any splitting of CFA.

Choose one of your RGB exposures as the Reference image.
Now stack your RGB exposures using that reference (e.g. use BatchPreprocessing)
Now stack your Ha exposures using that reference (e.g. use BatchPreprocessing)
Now stack your OIII exposures using that reference (e.g. use BatchPreprocessing)

Now you can combine the red channel from the Ha stack into the R channel of your RGB stack (e.g. use PixelMath)
Also combine the blue and green channels from the OIII stack into the B&G channels of your RGB stack (e.g. use PixelMath)

You just need to experiment with the proportions of each.

Mark

[pfile]

if you have used SplitCFA then there's no debayering necessary. what you are left with is the equivalent of one channel of a CR2 which has been superpixel debayered. the only difference between superpixel debayer followed by splitting the channels is that with splitCFA you get 2 separate green images. if you debayer and split, the green channel that you get is the average of the two green channels of the CR2.

rob