Combining Ha data to LRGB

[LD]

This might be slightly off topic, but I'll pull it back.

Someone on another group I frequent asked what is the best method to work Ha data into an LRGB image. I know that I've usually just shot Ha instead of the L and used it as my luminance. But other people seem to blend Ha with either the Red or the L in effect having five channels to play with.

My questions are two: does anyone have a preference for how to use Ha? And, if blending with another channel, how is it best done in PixInsight (every other solution I see references PhotoShop...you might have heard some imagers use it :wink:

Thanks for any input,
Larry

[avastro]

Hi Larry
I recommend you to read Rob Gendler's ( a top astrophotographer ) article about "Color Imaging of Nebulas using Hydrogen-Alpha Data".
http://www.robgendlerastropics.com/HARGB.html
You will get some very usefull informations.

Cheers

Antoine

[LD]

Antoine,
Thanks for that suggestion. But it also takes me back to my other question: what would be the best way to achieve that in Pix? How should I merge the Ha and R to achieve the lends he speaks of? (I'm one of those non-PS people--I do have Elements so maybe could try there but I'm really looking for a Pix solution.)
Thanks again for the article, and for any other thoughts,
Larry

[Juan Conejero]

Hi Larry,

(every other solution I see references PhotoShop...you might have heard some imagers use it

You're joking right? How's that possible? Come on tell me who's he (or she)!

:lol:

It depends on the image, but in general, using pure Ha data as the luminance isn't a good idea.

One reason is that the stars may be poorly represented in the final image if you do so, since they are usually much smaller in a narrowband image than in any of the broadband RGB channels. This happens with Ha, OIII or SII images.

More important is the fact that Ha is significant only for objects that emit strongly in that wavelength; mainly emission nebulas. If you use pure Ha as the luminance, and there are some reflection nebulas in the image, or other extended objects that aren't so red -which is frequent-, then your obtained rendition won't be correct.

Consider that the luminance supports image detail. So the luminance should be optimized to maximize the representation of image structures.

I think Ha should be mixed with red to form a more detailed and strong red channel. This should solve the problem with the stars, since the red image fixes the "holes" created by the smaller stars in Ha. Then the problem is how to use Ha to improve the luminance. This largely depends on the objects represented and on the desired results. Basically, I think this is a trial-error work to find the best/preferred blend between a clear luminance (or even a synthetic luminance obtained from RGB channels) and the Ha image.

In PixInsight, you have the PixelMath and LRGBCombination tools to perform these tasks. For example, to obtain a blend composed by a 30% of the Ha image and a 70% of the red image, you could use the following PixelMath expression:

   0.3*halpha + 0.7*red

assuming that the images have been assigned the 'halpha' and 'red' identifiers in PixInsight.

To learn how the LRGBCombination tool works, we have a specific tutorial:

http://pixinsight.com/tutorials/STD/LRGB/en.html

This tutorial doesn't use Ha because it's about a galaxy image (M63) and the original data included no Ha. However the Ha and red images, as well as the Ha and luminance images, if necessary, can be combined using PixelMath with simple expressions as the one above.

By the way, you don't have to care about the famous 'salmon pink' problem with LRGBCombination in PixInsight: with this tool there is *absolutely* no hue change when you perform a LRGB combination, so red stays red. Also, it allows you to boost color saturation without chrominance noise. This is explained in the tutorial.

Hope this helps. Of course, don't hesitate to ask anything about imaging topics here. If it's astrophotography or image processing, it's definitely on-topic!

[LD]

Juan,
Very helpful, thank you. It was that pixel math answer that I was looking for. I must say, I'd probably be less fearful of that application without the word "math" in there :wink: (if you saw my grades from High School algebra you'd understand).
Again, and as always, thanks for the help. And, yes, I'm sure I'll be back with more questions!
Larry

[vicent_peris]

Hi Larry,

I've several ideas for doing this... I say you the most simple I think:

As Juan says, the stars are bigger in the red image than the ones in the H-alpha image. But I think is a VERY bad idea to mix the H-alpha with the red for the nebula information. This is because you will have a very poor signal to noise ratio in the red channel compared to the h-alpha component.

So we need two things:

- Stars with red channels dimension
- Inmaculate H-alpha information!

Supose you have made ten minutes subexposures with both red an h-alpha filters. Ignoring the filter transmission specification, we will have basically the same amount of signal from the nebula in both images.

For the star size problem, we can combine both images with a Maximum operator (this operator will choose the maximum of each pair of values of both images). So as a result, we will have an image with bigger stars. BUT a part of the noise from the red image will be transferred to the information of the H-alpha nebula.

The easiest method I think now is simply multiplying the H-alpha image by a factor of >1. As the stars are very much dimmer in the H-alpha image, by multiplying this image you will end with a picture in wich the stars are of the desired size, but with the inmaculateH-alpha information from the nebula.


Well... good luck!
Vicent.


PD: Another recommendation:  Apply a minimum filter to the original RGB image to make the stars dimmer, to compensate as much as possible the difference of size with the H-alpha image.

[LD]

Vicent,
Thanks for your input. If I understand you correctly, you recommend just using the Ha as the red channel with some manipulation as you describe for the star size. And the L is just the L? So it would be an LHaGB image?
Looks like plenty of experimenting to do! (But in all cases maybe wise to shoot five ways: L, Ha, R, G and B to have the data to play with.)
I really think the fun of this hobby is that there are always so many ways to approach an image and thus always new things to learn. This site is a great place for that learning, and Pix really does offer the chance to try out many different approaches.
Regards,
Larry

[vicent_peris]

Yes Larry,

the L is just the L.  :lol:  You can obtain a "less" quality L from the RGB components itself. But the H-alpha component must go entirely to the red channel, I think, as H-alpha emission is purely red.


Good luck! Please, post here your results.
Regards,
Vicent.

[twade]

Don't forget to add a little to the blue channel, especially if you are going to replace the red channel with the h-alpha channel.  This may not be a necessary step if you are just adding a percentage to the luminance channel.  

Keep in mind, where there's h-alpha.  There's h-beta.  By adding a little of the h-alpha channel into the blue channel, you will prevent the "pinkish" look.  It's a little difficult to figure out how much to add to the blue channel since h-beta attenuates quite easily due to intervening gas and dust.  You should experiment with values between 10-30 percent.  The lower values would imply lots of intervening dust and gas.

Wade

[vicent_peris]

Good idea, Wade.

For doing this with the appropiate proportions, I think you can do two things:


- Buy an H-beta filter and make a short total exposures of the object with it. This image will not contribute to the final RGB, but only te know with precision the proportions of the H-beta / H-alpha components.

- Search for a spectra catalog where your object is in, and apply the proportions of H-beta / H-alpha it indicates.


Using the H-beta filters, perhaps you can find that the proportion varies along the object... I think you can then use a highly blurred version of the H-beta image to use it as a multiplyier of the H-alpha image in an instance of PixelMath.



Regards,
Vicent.

[LD]

Wade and Vicent,
I can see I'll need a few additional slots in my filter wheel :wink: . I've actually begun to play with all the LHaRGB combinations on some old images. Definitely a big difference from HaRGB. Much more natural and colorful and, whoa!, where did all those stars suddenly come from?
Thanks again for the discussion.
Larry

[LD]

This discussion really had me itching to do some rework on my recent Pac-man. Meager data to play with, but I did find the experiment fun and instructive. If you don't mind lots of scrolling, the results are here:

http://www.freewebs.com/drivewayastronomy/workspace.htm

If nothing else, it helped me hone my skills in Pix

Regards,
Larry Durst

[Carlos Milovic]

Hi Larry

Your new method seems to have produced excelent results in my opinion. Indeed a mix of the sinthetic luminance and Ha is much better. And this is specially true for star colors (and also for the background and nebula). Congratulations.

Just a small suggestion. Try different values on the RGBWorkingSpace. In other words, try using a non-standard luminance calculation. For this object, maybe RGB factor of 0.4:0.3:0.3 may be appropiate. Just try it, and have fun

[LD]

Carlos,
(You've been busy on the forums, I see :wink: ) Thanks for checking out the site and commenting. Indeed, the hardest part of the process is determining all the proper proportions for the channels, both on the Luminance Ha contribution and the color channels. You have to toss your standard color mix out the window it seems--I did everything here as even splits. I will definitely play some more (ice storm predicted tonight anyway!).
Regards,
Larry

[Jack Harvey]

Good improvement!  Mixing the Ha with other channels than red is something Chris Shurr has been advocating for some time.  He gave a presentation on his technique at the Advanced Imaging Conference in 2007  http://www.schursastrophotography.com/ccdgalaxies-ha.html