Another Ha Combine question

[vicent_peris]

Ok, let's go.

The basic idea es to decrease the brightness ratio stars/nebulas. If you see at both the R and H-alpha images, you will see that the fwhm is very similar. The only difference is that in the H-alpha image the stars are very much fainter. This is why, after raising the midtones in both images, the stars appear much bigger in the red image.

So the key is combining the red image with a multiplied version of the H-alpha image.

==========

The first step is to have the background level controlled. We will force a background level of 0.05 with a simple formula in PixelMath. We will extract a small preview that is perhaps representative of the sky background (in this image, I have selected a small rectangle near the left - top corner). Supose we are working on the H-alfa image:

Ha - Med(Ha_background)+0.05

Ha = the H-alpha image
Ha_background = the sky background area in the H-alpha image

We substract the median value of the sky background to the H-alpha image and then we add a little pedestal of 0.05. Remember to disable the Rescale checkbox.

We will make this operation with the other images (the R, G and B components), so the four images will have a sky background level of 0.05.

==========

We have to make two RGB images. The first is a normal combination with our broadband images. In this case, I have asigned simply a proportion of 1:1:1.

The second RGB image will be the enhanced H-alpha. Let's see how we are going to make this improvement...

We will multiply the H-alpha by a factor X, through this simple formula:

(Ha * X) - (0.05 * (X - 1))

The second part of the formula is simply to maintain the sky background level at 0.05. Remember to disable the Rescale checkbox.

After this operation, we will mix the H-alpha image with the broadband R component, through a maximum operation:

Max (Ha, R)

This will be our new R channel for the second color combination.

==========

This is the H-alpha enhanced when we raise the midtones level:



It seems really horrible, but don't panic. We are going to recover the correct chrominance from the first RGB image.

We will make the same histogram adjustment to the first RGB image and we'll extract the a and b channels of the Lab color space, with the ChannelExtraction tool.

Then we will insert these components in our H-alpha enhanced color image, with the ChannelCombination tool. You can see the result in my last message.

Of course, this image is not lineal at all. But you can return the image to a lineal state with a second midtones adjustment of one minus the first one.

That's all!
Regards,
Vicent.

[vicent_peris]

I forgot to say that you can change the RGB working space (with the RGBWorkingSpace tool) with a Gamma value of 1.0, otherwise you will get the classical pink and unsaturated H-alpha enhanced color image. Make this with Photoshop.  :wink:


Regards,
Vicent.

[vicent_peris]

One note more... I have been making some refinements to this technique, resulting with much more accurate color representation. We will implement this refinements in the script... For the moment, the final composite:





Vicent.

[Jack Harvey]

I have gotten all the way to the combine of the two RGBs (one without Ha and the other with).  The channel extract of a and b from the RGB without and combination of a nd b channels to the RGB with Ha does not give me a good image.  I get blue stars etc? So either I screwed up in all the math (don't think so) or there is more to combinging the RGB with the Ha_RGB.

Or I wait for the tool to do this??

[vicent_peris]

Don't worry, Jack, I'm writing a small tutorial, I hope to have it this night. I know the instructions I gave yesterday are demanding some screenshots. By the way, I'm refining the method...

See this new image... It's the result of applying a H-alpha gain of x20. Of course, the stars are somewhat bizarre, because you don't have sufficient luminance support for them, but the image is IMHO really spectacular:




We are going to do a module for PixInsight in the next weeks, inmediately after the launch of PixInsight. This module will not only integrate H-alpha with RGB, but other narrowband filters too, like O-III, H-beta or S-II. Be prepared.  :wink:



Vicent.

[LD]

You mean just when I'm beginning to get a grasp on the complexities, you're going to make it easy? :wink:

This continues to be an interesting topic with astounding results, and once again shows the great possibilities of PixInsight.

Regards,
Larry

[Jack Harvey]

A module to integrate narrow band into broadband would be unique in the world of processing software.

[vicent_peris]

Finally seems that the processing required is harder than I thought at the first moment, so please be patient... I think we will release two separate tools. One for generating a map of broadband emiting objects in the image, and another one to integrate broad and narrowband images.

Vicent.

[Jack Harvey]

I am still playing with this also.  BTW in the first step where you control the background using a small part of the sky which you select as a retangle, could you not just use ABE and then

Ha-Med(Ha_ABE)+0.05

[Jack Harvey]

Vincent I posted my fianl image using this technique on the PixInsight ftp in folder jack.harvey.  THis is a promising technique

[vicent_peris]

Hi Jack,

you ABE idea looks very good. And can be better for making all automatic. Thanks you!

My problem now is noise transference from the R image to the H-alpha one. In your image, the H-alpha signal in the R image is nearly as good as it is in the H-alpha image. But if you add some noise to the R image (PixInsight has a module for adding noise), H-alpha will be degraded... For the moment, I have not a good solution.  :cry:


Vicent.

[Jack Harvey]

Hmmmm  Not sure I have a quick idea on this one?  Buena Suerte!

[vicent_peris]

At least, if you have good enough red data, this method gives you a way to raise the nebulas over the stars.


Vicent.

[Jack Harvey]

Actauly I do have new Red data that I got last night.  In fact it was a great night and I got all new data.  I can leave you on the ftp the new data under a folder named New 7024 IF you want it.  It will be there.

[vicent_peris]

Ok Jack, this would be great. One more thing... could you upload a R image with bad S/N ratio? I need it for experimenting.

Now, I think I have the solution... This is the RGB with noise added to the R channel:



And this is the new HaRGB:



As you can see, now there isn't any noise transfer from the R to the H-alpha signal.  :wink:


I hope this will work with other images... But now it's a pretty complicated technique, so I think can be better to wait for the module...



Thanks you a lot, Jack.
Regards,
Vicent.