Yet another HaRGB question

[Ioannis Ioannou]

Hello,

I'm a bit new on this, and I tried to read any previous threads before posting.
Except of Jack Harvey's Technique http://pixinsight.com/forum/index.php?topic=562.0, which I still have to figure out the details,  I think I've tried all the other suggestions to combine my Ha data with RGB.  I got the best results using Gendler's 2nd technique, ie I used the (1-k)*$T+k*Ha formula in pixelmath and combined many times Ha+R , starting from 20/80 and going up to 80/20  with steps of 10% increase each time. At least this gave me an improved RGB image with good saturation - matched with the original RGB data, reach star field, but not all of the detail contained in Ha.

At this point combining Ha fairther with PixelMath gains nothing, so I tried some other ways, ie LRGB with Ha as L and many combinations, extract L from the improved RGB image and combine it with Ha , replacing L with Ha (I guess this is the same as LRGB ?) etc
But the results are not good: I may get better nebula's details but I'm loosing the star-field, the colors (I guess this is matter of correct match in saturation) or the shift (in some cases I'm getting the infamous salmon). Any ideas? Actually keeping the star-field and replacing the nebulosity only will be enough, maybe a star mask ?

[Nocturnal]

Hi,

this question comes up all the time and I feel compelled to answer in the same way. Why would you try to make Ha the L channel? Think about what you're doing. You're modifying the intensity of *all* colors based on how bright Ha is. So if there's a region with lots of green but no Ha then the green will disappear. This makes absolutely no sense to me, sorry. Sure it can be done but what's the point?

The only thing that makes sense to me in this case is to replace R with Ha or to create an Ha image in some arbitrary color and layer that on top with some amount of opacity so it lies 'on top' of the RGB.

My advise: make monochrome Ha images (they're often wonderful) or add some other NB filters and combine those with Ha to create true NB images ala Hubble. The whole Ha+RGB thing is overdone, IMNSHO (In My Not So Humble Opionion)

If you're dead set on combining Ha and RGB then simply formulate what it is you want to achieve (eg "replace red with Ha" or "make green dimmer where there's no Ha" etc) and you'll quickly know how to achieve that. There is no magic formula and blindly trying things probably won't get you very far. Try to understand what you are combining.

I hope I'm not a downer. Sorry if it came accross that way.

[Niall Saunders]

Another possibility is to put a little Ha into the Gn channel as well. If you have already used 'k' to blend the Ha with Rd (using your formula), then also try 'k'=(1-k) in the same formula (or, indeed some other empirical value of 'k' altogether) to blend a little Ha into Gn, and then rebuild the RGB image.

I agree with Sander - there is not a lot of mileage in using Ha in the L channel.

You can also use Masks to decide 'where' you want the Ha enhancement to be applied.

Just keep in mind that you are in the 'artistic zone' here - whatever you feel 'looks good' is then 'good enough'.

Cheers,

[Ioannis Ioannou]

Thank you for the replies. Since I'm new on this field, and my theoretical background is minimum, my only resources is Internet, other people's opinions and trial-and-error. The reason of course I try this (beside that I can see other people doing - or trying - it) is "cheating" over non-optimal conditions (equipment, light pollution, moon etc). If I get this straight, there is no benefit trying HaRGB (ngc6888 will take too much time with LRGB, not to mention the lack of time and dark site), and better to focus in NB imaging in cases like this, right ? Although I must admit that I was impressed applying the technique I mentioned, indeed it gave boost to my RGB data.

Thank you

[dhalliday]

Here is a recent OIII and Ha version;
http://www.flickr.com/photos/daveh56/4755807228/

I still think your question is an excellent one !!\
Wish I had an answer.
Just as a comment though...I think NB is the way to go (or focus) if your sky sucks anything like mine.

Dave
http://www.flickr.com/photos/daveh56/4755807228/

[Nocturnal]

Hi,

if you don't have dark skies and RGB or LRGB isn't feasible then HaRGB probably won't make much sense either. In that case I'd do just Ha or get more NB filters to add 1 or 2 more channels. Still you need to pick your targets carefully, not all of them emit all NB wavelengths in large quantities. In other words just because you put a filter on doesn't mean you'll get a picture

Niall: I don't see why you'd 'blend' (whatever that is) Ha into green. I've heard that before and I don't understand it. So now in areas where there is lots of Ha you get lots of Red and Green rather than just Red. The original green signal gets changed and no longer represents green wavelengths coming from the target. How does that convey extra information?

[Simon Hicks]

Sander, the reason for blending some Ha into the green is that hydrogen emits at the Halpha (656nm) and the Hbeta (486nm) wavelengths in a roughly constant ratio. So if you capture some Ha at 656nm then you know that exactly the same structure is actually present at 486nm....to a first approximation. So measuring the 656nm structure does actually tell you what some of the 486nm structure will be. There are plenty of other wavelengths, plus a continuum as well...but these are less intense.

[mmirot]

I have seen most people add a fraction to Blue rather than green. Typically 85-95 of Ha is applied to red and 5-15% to blue.

Max

[Simon Hicks]

Hi Max,

I think you are right.

However, I think 486nm is sort of on the cusp between a blue filter and a green filter. I think they overlap a bit at this point. In other words if you fired 486nm light at a colour DSLR then both the blue and the green pixels would give a response....with the blue giving a larger response than the green.

So this suggests that the Ha image should be blended something like 85% into the red, 12% into the blue and 3% into the green (just guessing at the figures). But maybe the green isn't worth the effort?

Cheers
         Simon

[Niall Saunders]

Sander,

Don't get me wrong, I do agree that you can't make a silk purse out of a sow's ear, as we say here, but there are no 'rules' when it comes to experimenting. And that is what we are talking about here. Trying to get an 'aesthetic' image from limited data. I am even happy to call this 'artistic' image processing, not 'scientific' processing - although it is still nice to see the processing stages avoid the 'paintbrushed masks' so often used in a PS environment.

I also agree that, if you can only get reasonably clean' data through a single (Ha) NB filter, then why not just present that as a Mono image? Why try and incorporate poorer-quality RGB data into the blend, and then have to 'fight' to build a 'nice' image afterwards. Similarly, if you then resort to full tri-channel NB imaging, which target objects are going to give you sufficient signal to fill the other two channels?

In these circumstances, object selection and acquisition methods are always going to be a compromise, as is image presentation. All that can be offered, by way of advice, is suggestion - i.e. suggesting 'try this', or 'try that' - which is where I think we are all coming from. There is no 'right way', other than - perhaps - your suggestion to 'keep it simple stupid' (KISS) and just make the 'best' monochrome image out of the single-channel NB data.

Sometimes though, you just want to see if you CAN 'improve' things. Sometimes even a 'leather purse' might be better than 'just a pig's ear' 

Cheers,

[Jack Harvey]

Chris SHur has done lots of work developing Ha combine techniques.  He presented his newest at AIC in 2007, Enhanced Hydrogen Imaging.
An overview is as follows:
  
1.  First make a standard LRGB
2.   Take additional Ha frames
3.   Subtract Red data from Ha data
4.  Add difference to Red, Green, Blue
5.  Recombine color image

Step by step instructions and example images in this PDF:

http://www.aicccd.com/archive/aic2007/SchurAIC2007.pdf

Buena Suerte Amigos!

[Nocturnal]

I dunno Jack. I took a look at the presentation and I'm a little surprised you're endorsing it. What he's doing is completely arbitrary, no different from painting. He has Ha data which not surprisingly looks a lot like the L data. After all he's imaging a galaxy which means mostly stars, all of them emitting Ha. He then subtracts Red to get 'Hydrogen', whatever that is. He then adds the result back to Red. What does it even mean to subtract Red from Ha when Red already contains Ha? YOU GET NEGATIVE IMAGE VALUES! He doesn't even acknowledge that or explain what PS does in that case.

Red = (Ha + other red data)

Ha - (Ha + other red data) = - other red data

Supposedly the negative values loose their sign and become positive again, not sure.


I think if he simply used the Ha data as a mask and did a saturation curve he'd get very similar results but it could be justified and explained.

[mmirot]

It is a good techinque Vicent even uses his own method which is more complex. He will go into detail at Adler.

http://pixinsight.com/workshops/adler-2010/ 

[vicent_peris]

Hello,

my idea when processing M51 was to independize H-alpha emission from broadband R emission, because there's a lot of broadband emission contaminating the line emission in the narrowband image. I derived some equations that do this job. Knowing the bandwidth of our filters, and assuming that exposure times and atmospheric extinction are the same, we have:

C = (Br * (R - Ha)) / (Br - Bh)

L = ((Ha * Br) - (R * Bh)) / (Br - Bh)


where:

C = broadband continuum emission

L  = line emission at H-alpha

R = broadband R image

Br = bandwidth of R filter

Ha = narrowband H-alpha image

Bh = bandwidth of H-alpha filter


This is very easy to apply with PixelMath (as you will see at the Adler). The H-alpha image before continuum subtraction is below:



And the continuum cleaned H-alpha image is this one:



Once we have the cleaned H-alpha image, we can ADD this to the R filter, multiplied by a factor. In our H-alpha enhanced version of M51, H-alpha emission was multiplied by a factor of 4.

All this must ONLY be done when your images are linear. Otherwise, you cannot subtract the continuum and be sure all that faint hydrogen structures are real in all the galaxy body.


Hope this helps!
Best regards,
Vicent.

[Nocturnal]

Hi Vicent,

your formula assumes that broadband R intensity is evenly distributed across the entire range, correct? Once you have estimated the actual Ha signal you're making the image more red in those areas that have Ha.

It sounds like Chris is trying to do the same thing but not doing a very good job explaining it. As far as I can tell there is no normalization for the R subtraction based on pass bands of R and Ha filters. Perhaps it was in the talk but this is so essential that it should have been in the slides as well.