StarNet with Star Mask to preserve star colors

Bruce Donzanti

I tried StarNet several times with so-so success. I tested it on two images- Crescent nebula and the Needle Galaxy. The difference for testing purposes was mainly because the Crescent image has a ton of stars and the Needle Galaxy has few. So, I understand to use StarNet early on but I used it only on a mildly stretched image in an attempt to preserve the star color and create a star mask to add the stars back later. So far, this was fine in both cases. I then continue to process the starless images and finally add the stars back in using PixelMath. In the case of the Needle Galaxy it worked fine. The Crescent nebula was a disaster because the starless crescent still had a ton of star remnants that came out like ugly noise in the background. I attempted to use CloneStamp to clean it up but that did not work well. So, I am obviously doing something wrong. Any suggestions?

BlockHead

A couple of notes.FIrst...we are talking about permanently stretched non-linear images wwth Starnet.

Typically the artifacts that Starnet leaves behind are in star locations... so that when you add the stars back they get covered. It is possible to enlarge/enhance the artifacts when you process the starless image.

Secondly... (no one listens to me on this topic ... lol )

I don't add stars back using PixelMath... not in an additive fashion. I think there are better ways (and I demonstrate in Horizons). No masks... no math...no muss. I blend the stars in. Typically with Screen blending mode. This is an "Adam Block" technique that I find pretty powerful. 

I guess I would need to see an example area... you can attach jpegs to this forum now! Makes things easier.

-the Blockhead

Bruce Donzanti

Hi Adam

This is Bruce, the guy you spoke with over the phone about my Crescent Nebula.  I guess I am the blockhead and not you.  I believe I posted this prior to us chatting.  Anyway, here is my Needle Galaxy.  In general, I like my images but I still struggle with preserving star color despite the stars not being saturated.  I have tried using PCC and ArcsinhStretch and they seem to help a little.

Thanks for any additional input.

BlockHead

Well... honestly that field doesn't have that much variation in star color. 

Just remember to protect the stars if you do noise reduction (especially on the chrominance.)

-the Blockhead