Techniques for blending RGB stars into narrowband images

[MikeWiles]

I've attempted half-heartedly to do this in the past and met with unsatisfactory results.  I'm working on an image now where I'd like to give it another shot - blending regular RGB stars into a narrowband image.  I'm planning to create a synthetic RGB images from H-Alpha and OIII data because the luminance has so much detail and contrast that tends to get lost in a regular broadband image.  The stars in this image will be plain white as a result of the narrowband data.  I'd also like to try this with a Hubble palette image.  What technique do you use to get RGB stars into the image and have it look realistic?  I hope my question makes sense.

Thanks,

Mike

[topboxman]

Your question makes perfect sense. I would also like to know as well.

Peter

[Geoff]

The NBRGB script seems to work quite well to do this.
Geoff

[topboxman]

Do you mean this one?

http://pixinsight.com/forum/index.php?topic=3446.msg40130#msg40130

Thanks,
Peter

[Geoff]

Yes, it started out as an HaRGB script, but round about page 7 morphed into a more general script for adding narrowband to RGB, hence the (N)arrow (B)and RGB.
Geoff

[Phil Leigh]

I'm finding with the NBRGB script that the Ha stars and RGB stars don't combine nicely because they are different sizes and colours... the bigger and brighter the stars are, the worse the problem is. I end up with white stars with a red centre or sometimes a red edge.

[topboxman]

Me too. My stars are also white when using NB image with RGB image using NBRGB script.

I think what Mike is asking is to replace NB stars with RGB stars to make stars look more natural without affecting the rest of NB image. I am not sure NBRGB script can do this.

I have seen many NB images (Sii Ha Oiii mapped to RGB) with RGB stars using Photoshop. Mike is asking how to do this with PixInsight.

Peter

[marekc]

topboxman said:

"I have seen many NB images (Sii Ha Oiii mapped to RGB) with RGB stars using Photoshop. Mike is asking how to do this with PixInsight."

I'd love to know how to do this in PI, too.

- Marek

[Phil Leigh]

Alternatively, is it possible to completely remove the stars from the Ha image?

[topboxman]

I think you can do that with Morphological Transformation. It's an iterative process. Star Mask then Morph then new Star Mask then Morph and keep going until stars are gone. It needs to be done in linear.

http://www.harrysastroshed.com/linmask.html

Peter

[Phil Leigh]

I think you can do that with Morphological Transformation. It's an iterative process. Star Mask then Morph then new Star Mask then Morph and keep going until stars are gone. It needs to be done in linear.

http://www.harrysastroshed.com/linmask.html

Peter

Brilliant - thanks Peter.
I used an inverted linear star mask to select the largest stars on the grayscale Ha image then repeated iterations of Morphological Transformation until they were no longer visible. That allowed the Ha to be blended into the RGB using the NBRGB script without any horrible star artifacts.
I wish MT had a dynamic preview... (I wish all the tools had a dynamic preview!)
regards
Phil

[topboxman]

Nice!!!

Great idea about removing stars from NB images. Never thought of it.

Thanks,
Peter

[MikeWiles]

I think what Mike is asking is to replace NB stars with RGB stars to make stars look more natural without affecting the rest of NB image.

This is/was exactly the question. 

[Phil Leigh]

Mike, if you are using a synthetic RGB built from just Ha and Oiii the stars are never going to have the correct/realistic colours.

Unless I am misunderstanding you?

What I would do is blend your Ha and Oii data into a true RGB image, losing/minimising the Ha and Oii stars and replacing them with the RGB versions.

I guess this will only work if you have a true RGB image as well as your narrowband data.
Regards Phil

[RickS]

I just finished processing a Hubble palette narrowband image of M 16 and had some success with blending in RGB stars.  Here 'tis:



I created a star mask, applied it to the NB image protecting everything except the stars, and then used PixelMath to assign the values from the RGB image to the NB image (both images were stretched first).  Of course, it wasn't quite that simple in practice.  I had to create the star mask quite carefully.  I also made sure that the brightness of the stars was similar in both images and went to some trouble to keep the stars under control in both images with deconvolution and several applications of MorphologicalTransformation.  But it wasn't impossibly difficult and I think the results were worthwhile.

Cheers,
Rick.