Combining RGB + Ha + OIII

[ramv]

Hi,
Is there a theory paper/tutorial on how to combine RGB +ha +OIII data?
Basically I have binned 2x2 data for RGB, binned 2x2 data for OIII and unbinned Ha data for NGC6888 that I wish to combine into a color image. I saw Vicent Peris' paper on combining Ha+ RGB and understand that there are scripts that codify this theory.
Is there something similar that can be done to add OIII to the resulting image? The challenge I see is that OIII is not a pure color (while Ha can be mapped to red) OIII channel has the same set of issues/challenges as Ha with star sizes/brightnesses being different from RGB atleast in the non linear stage.

Any advice would be much appreciated!

--Ram

[Geoff]

Although OIII has emission bands at 496 and 501 nanometres, these are both in the green part of the spectrum, so you could try combining it with the green channel in the Vicent Peris way.  I've got no idea whether or not this will work, but it's the first thing I'd try.
Geoff

[ramv]

Geoff, I might be mistaken but according to astrodon, OIII is a teal color (something like a blue-green mix).  I was wondering if it was possible to precisely mix this color with RGB.

Thanks,
--Ram

[Geoff]

I think "precisely" is the wrong word to use here. When you mix NB and RGB you are trying to get a good looking result-- low noise and plausible colour. Teal is close enough to green to try what I suggested.  If it works, that's good.  If not, then other options have to be considred.
Geoff

[DavidW]

Hello, I'm on day 3 of my trial and have been focused on learning the program with this goal in mind. I have an idea for workflow on how to do this and would like to see what people think. I have never used Photoshop but have used Paintshop Pro for levels and curves on a few DSLR images, I couldn't align layers so looked here - anyways I'm pretty open to new workflows.

Test Color Hues
- from the menu, select Image-->New to create new image with RGB colorspace, use default initial values
- for Ha try these values and click 'Apply Global'
R - 0.81
G - 0
B - 0.23

- for OIII try
R - 0
G - 1
B - 1

I was thinking of using PixelMath to add the grayscales of the OIII and Ha to the RGB but the OIII kind of threw me for a loop so I started thinking I should create RGB color images of the narrowband using these proportions so I've started experimenting with this path.

Last weekend was my first time out with my QSIwsg and so I cheated a bit during image acquisition with Maxim and applied Simple Darks (no flats, no bias) but did collect data for 6 different filters (L R G B Ha OIII) of M27 and the Veil.

StarAlignment and ImageCallibration was done in PixInsight and I have to say I really like how StarAlignment and ImageCallibration work (and thank you Harry for the video tutorials). During StarAlignment I used the same luminance image sub as reference for all of the filter subs and it scaled the Red subs automatically for me (somehow binning got set to 2 for the Red subs).

The Processing Workflow:

Create linear RGB color Image

1) LRGBCombination of Red, Green and Blue stacks
2) BackgroundNeutralization
3) ColorCallibration to set the white balance
(anything else that should be done at this point?)

Create Ha and OIII RGB images
1) from the menu, Image-->New to create new image with RGB colorspace, change width and height as required to match camera, use default initial values (0), and 'Apply Global' to have an active RGB window

2) PixelMath to add the narrowband grayscale stacks to the channels of a new RGB image. Have 'Create new image' checked and click 'Apply'. Use ScreenTransferFunction and autostretch without 'Link RGB Channels'

for Ha
R - (1-M27_Ha)*0.81
G -
B - (1-M27_Ha)*0.23

for OIII
R -
G - (0+M27_OIII)*1
B - (0+M27_OIII)*1

This is where I am at right now.

I guess BackgroundNeutralization could be done on the narrowband RGB images now. Not sure what should be done for these narrowband images. Would white balance still apply?

I was thinking I could maybe register and stack the RGB's but now I'm thinking  PixelMath to blend red channel to red, blue to blue, green to green might be better but haven't thought it through regarding ratios.

Any thoughts on what I'm doing?

David

[DavidW]

I think that PixelMath formaula to create the narrowband RGB for Ha needs to be correted, the OIII I think is ok.

for Ha
R - (0+M27_Ha)*0.81
G -
-1
B - ((0+M27_Ha)*0.23)-1

for OIII
R -
G - (0+M27_OIII)*1
B - (0+M27_OIII)*1

Now I have black background with hued data.

David Weiss

[Philip de Louraille]

Read this thread and see why the weights need to be consistent.
http://pixinsight.com/forum/index.php?topic=4135.msg28796#msg28796

Notice that in all cases, the sum of the weights in the RGB are all equal (in this case they all sum to 1.0 [normalized to 1])

[DavidW]

Thank you for replying Philip,

What started me on this path was a presentation by Debra Ceravolo at a star party last year, it is just recently I got data to work with to try it out. Here is a link to what her presentation was about.
http://adsabs.harvard.edu//abs/2011SASS...30...13C

I'll also mention at this time that her presentation mentions use of a 'Screen' blend, I found this link for PixelMath formulas to mimic Photoshop blending modes.
http://blog.deepskycolors.com/archivo/2010/04/21/formulas-for-Photoshop-blending-modes.html

Found it.
Thread: http://pixinsight.com/forum/index.php?topic=2513.msg25601#msg25601
Juan recommended this combo for RGB:
0.5*S2 + 0.5*Ha
0.15*Ha + 0.85*O3
O3

I do like this blend for narrowband data, it does look good and looks like it uses some Ha to represent Hb too but I think the OIII hue would be off by 15%. In trying to be more accurate what I did in my previous gives one individual RGB images for HA and OIII and balances the data in the proper channels to get the desired color. I have also used regular broadband filters and created a RGB image for those so for my final image I will want to combine 3 RGB images and a luminance and after thinking about it for the last day I think the way to approach it is to use PixelMath and do a straight addition of the pixels values and do use the rescaling checkbox to bring all of the pixel values back into bounds.

I still need to understand using the rescaling option better though, what I would like is a consistent across the board scaledown after all the addition. I might need a PixelMath function but I hope not. Using the addition of the OII and HA makes me feel better that the intensity of the gases is preserved, adding the broadband RGB signal is more like the cosmetic touch.

I think this might complete the workflow I was looking for to 'get a picture', but I still need to learn a whole lot about the image processing processes that I think are the heart of this progam - the wavelets, transforms and other stuff that helps one smooth images, bring out details and reduce noise.

Anybody else have thoughts on this?

David Weiss

[DavidW]

Actually straight addition of pixel values might be a bad idea, this might be better:
RGB/K = (RGB*0.34) + (OIII_RGB*0.33) + (HA_RGB*0.33)
into a new RGB image

I've already figured out the channel values for each image so this should be all that is necessary. No scaling this way either. Then add the luminance.

[pfile]

thanks for researching this and posting it. i think it will be very useful.

[ramv]

David,Thanks very much for the research and links to the paper.
This is exactly the kind of information I was looking for. I will have a go at this as well with my data.