Blending narrowband

[Boyte1217]

Is there a tutorial, that will show me how to blend narrowband like Ha and OIII together in Pixinsight.

[Juan Conejero]

Hi,

PixelMath is the tool of choice to perform this task in PixInsight. For example, say you have three narrowband images: Ha, O3 and S2. The first thing you have to do is deciding the palette you want to use. If you are going to use the Hubble palette then the process is straightforward: you just have to map S2 to red, Ha to green and O3 to blue. Do the following:

- Open your three narrowband frames. All the images must be co-registered and should have the same dimensions. StarAlignment is the tool of choice to register all kind of deep-sky images in PixInsight.

- Set image identifiers. Bring the Ha image to front, select Image > Identifier and type Ha as the new identifier, then click OK. Do the same for the other two frames to set their identifiers to S2 and O3, respectively. Note that instead of selecting Image > Identifier you can also double click the vertical tab at the left edge of each image window.

- Open the PixelMath tool (from ProcessExplorer for example).

- Disable the "Use a single RGB/K expression" option.

- Enter the following expressions, respectively for the R, G and B slots:

S2
Ha
O3

- Disable the "Rescale result" option.

- Open the Destination section of PixelMath and select:

* Create a new image
* Color space = RGB Color
* Sample format = 32-bit floating point (this is optional, but advisable)

- Now execute PixelMath on one of the images (it doesn't matter which one because we are going to create a new image). Drag the blue triangle to the image and wait a few seconds.
 
This will generate a combined RGB image with a pure Hubble palette. Personally, I don't like the Hubble palette for aesthetical reasons. A palette that mixes S2 and Ha for red and renders O3 as cyan has given me very good results with many images. For example:

0.5*S2 + 0.5*Ha
0.15*Ha + 0.85*O3
O3

where the different factors must be fine tuned for each particular case. In general (unless one of the frames is very weak), the sum of factors for each channel must be one; otherwise you'll be saturating bright regions. By mixing channels in this way, you can implement quite sophisticated compositions.

I hope this will give you enough guidelines for narrowband compositions with PixelMath. There are other ways to use PixelMath for this task. For example, instead of creating a new image you can work on a previously generated RGB image, where you can define previews for fast trial-error work. Then there are other techniques for RGB+NB compositions, which can also be implemented with this tool. But let's see one thing at a time.

[Juan Conejero]

Welcome to PixInsight Forum, by the way

[Boyte1217]

THANKS Juan for the info, this is exactly what I have been looking for!

WT

[Philip de Louraille]

Choosing how to blend narrowband is a difficult task.
On one hand, the data obtained is the direct result of local physics (as opposed to mostly hot gas showing off its blackbody spectrum [plus the emission and minus the absorption lines] that a regular RGB dataset shows.)

So the colors for a narrowband dataset are artificial. I teach a lab where I ask students to experiment with what color to assign Ha, S2 and O3 to show the most "contrast" between the colors so as to better see them.

The mix Juan shows above does indeed do a very good job at getting the narrowband to show colors we are use to; but some details are getting lost due to low contrast (it is hard to beat yellow/green/cyan against black...)

I was just working on such a dataset the last 3 days and was writing a reply to this request but Juan beat me to it. )

The first pic is the Hubble color set (S2, Ha, O3) and the second is Juan's mix.

[Juan Conejero]

Hi Philip,

Excellent comparison. I really like both versions. Palettes that map each band completely into a particular RGB channel —Hubble is just one of the six possible combinations— provide the best visual separation of the data. For aesthetical purposes though, I tend to prefer more "natural" palettes. Of course, we know that the concept of natural color doesn't make sense in astrophotography, and much less in NB imaging, but Ha and S2 emissions are at the red side of the spectrum and O3 is closer to a teal color than to blue. This is the fun of narrowband imaging, that you can choose whatever palette you want because everything is arbitrary regarding color

[Josh Lake]

Juan, I was doing a search on Narrowband combinations and ran into this old thread. I was wondering if you could expand on the 6 possible combinations that give best color separation. I've become a real expert at using PixelMath to blend my narrowband data, but I'm still a novice with bad instincts when it comes to actually creating a good palette. I loved the 'Hardcore' rosette that was posted here and was wondering if there were other good combinations to try.

I recognize that it is essentially a trial and error process. I'm will to try, no doubt about it, but I always question whether I am missing some great combo just by chance.

Also, do you know of any further progress on that script that would allow us to explore color combinations on the fly with a real time preview? I will keep searching for that thread.

Hi Philip,

Excellent comparison. I really like both versions. Palettes that map each band completely into a particular RGB channel —Hubble is just one of the six possible combinations— provide the best visual separation of the data. For aesthetical purposes though, I tend to prefer more "natural" palettes. Of course, we know that the concept of natural color doesn't make sense in astrophotography, and much less in NB imaging, but Ha and S2 emissions are at the red side of the spectrum and O3 is closer to a teal color than to blue. This is the fun of narrowband imaging, that you can choose whatever palette you want because everything is arbitrary regarding color

[neilfleming]

Juan, since at 486.1nm, the H-beta component is on the blue side of the OIII line, I've always incorporated the H-beta proxy from the H-a data into the blue channel.  Why did you choose the green instead?

   Cheers...Neil

[Juan Conejero]

Hi Neil,

Why did you choose the green instead?

Just because this palette looked very nice when I used it for a NGC 5189 image Then I applied variants of the same idea for other images, also with nice results IMO. My intention in this case was not to use a small part of the H-alpha component as an emulation for H-beta, but just to fine tune the rendition of OIII as a teal color (cyan), for purely aesthetic reasons.

Adding a small part of H-alpha to the blue channel to represent the ratio of H-alpha to H-beta emission is also a good option; probably better than mine because it has a physical justification, if that's the criterion you want to apply.

[jerryyyyy]

Has there been anything new in terms of PixelMath combinations in narrow band since this thread went cold?  What about combinations just of O-III and H-alpha?