Ha + RGB clarification

[strivedi]

Hi,

I am trying to learn effective ways to combine Ha with RGB for M51 image that I captured in June. 

I ready Vicent's article and tried developing a synthetic broadband image from a continuum map (C = Red/Ha, then noise reduce C and RedSynthetic = Cnoisereduced*Ha) then generate iterative RedSynthetic images using a different continuum map but the same Ha image.

For the life of me my continuum image shows that is is completely white. This makes mathematical sense since Red divided by Ha will give a relatively large number close to 1.0 for each pixel.

But Vicent's example shows a very clear continuum map which I am unable to get. 

Next I tried the method outlined by Harry on his tutorial to create a new Ha with a forumula like HaNew = (100*Ha - 10*Red)/(100 -10) and then recombine it with Red using a forumula such as New Red = $T + (HaNew - Med(HaNew)*Multfactor.

Playing with the 100, bumping it to 230 for my image and Multfactor or 20, I get acceptable HaRGB results where the HaNew has almost no continuum and the HaRGB looks good.

My question is - why does Vicent's method give me an all white continuum map? What am I missing?

I also tried the NBRGB script but for the life of me I am unable to get rid of the all blue color of the resulting image.

Shailesh

[troypiggo]

I was playing around with this very thing (again) last night, and found exactly the same problem.  Red/Ha is very bright.  I also struggle figuring out how the iterations work, because taking the new RedSynthetic (as you call it) and dividing again by Ha gives you exactly the same continuum C as you just had - unless I'm missing something in that step.

I have seen this question asked many times, and have asked it myself several times, and have never seen a conclusive clarification on the method.  Invariably you just get pointed back to the NBRGB (formerly HaRGB) scripts.
http://pixinsight.com/forum/index.php?topic=4120.msg28692#msg28692
http://pixinsight.com/forum/index.php?topic=3446.msg29566#msg29566
http://pixinsight.com/forum/index.php?topic=5644.msg38594#msg38594

Where I ended up last night was exactly where you did with that PixelMath solution with nonlinear (stretched) images:

R=max($T,k*(_Ha-($T-_Ha)*5/100))
G=$T
B=max($T,BAL*_Ha)
k=0.9,BAL=0.15

The 5/100 is for my Astrodon filters.  The BAL is to put some Ha in B for bit of balance, but not necessary.  k is a factor to play with for balance too.  The formula for R was from here: http://pixinsight.com/forum/index.php?topic=2298.msg15353#msg15353, but don't believe it is truly faithful to "Vicent's Method" as I understand it.

The results were pleasing enough, and it was late, so I left it there for now.  Would like to investigate further if anyone can shed some light here.  This has to be one of the most common questions - how to blend Ha (or any NB) with RGB - and "Vicent's Method" gets referred to a lot, yet haven't seen any further clarification or posts on it since the original was written.  I'd be happy to be proven wrong and pointed to posts that do.

[troypiggo]

...
Next I tried the method outlined by Harry on his tutorial to create a new Ha with a forumula like HaNew = (100*Ha - 10*Red)/(100 -10) and then recombine it with Red using a forumula such as New Red = $T + (HaNew - Med(HaNew)*Multfactor.

Playing with the 100, bumping it to 230 for my image and Multfactor or 20, I get acceptable HaRGB results where the HaNew has almost no continuum and the HaRGB looks good.
...

Just tried the above.  How does the "...Med(HaNew)" part work?  Doesn't PixelMath's med() require 3 or more images?

[strivedi]

I believe the Med function returns the median value of the argument. So if my understanding is correct, you are removing the median pixels of Ha from Ha.

Based on your feedback and one of the links provided I tried HaNew = Ha - (Red - Ha)*0.07  and then tried to color combine with Red in 60/40 ratio.

I compared it with my previous method per Harry's tutorial that I had mentioned earlier, and I found that that method gave better Ha globules in the M51 arms versus this method.

Shailesh

[oldwexi]

I believe the Med function returns the median value of the argument. So if my understanding is correct, you are removing the median pixels of Ha from Ha.

Shailesh

No you are not removing the median Pixels you are simply only using the Pixels with a value higher than median.
As the median value usually is in the Histogramm the tip of the Histogramm shoulder which represents
most of the background,
So, by subtracting the median value from each Ha pixel you use only the object Pixels and not the Background Pixels
and add them to the red channel.
Gerald

[strivedi]

Got it. Thanks for the explanation Gerald.

Shailesh

[strivedi]

I came across an excellent tutorial on how to use Vicent's method for Ha RGB combination from Colin McGill at this link: http://www.astrobin.com/68513/0/

I see that his M81 image really pops with Ha globules, which is what I want to aim for in my M51 capture.

This is very similar to Harry's tutorial.

I will try it out tonight.

Shailesh

[jerryyyyy]

In PI under scripts/MultiChannel Synthesis there are the AIP scripts one for Ha-RGV (V=verte=G=green). 

You might try this at the default settings, then you can manipulate various mixtures ... would be interested in seeing some images posted with comparisons of methods.