It's a bit painful to do but removing all the stars from the NB images for the colour combine produces really nice results, Colin. That's the real core of JP Metsavainio's tone mapping technique. Now that you're a ninja master at star masks it should be easy. I mask the stars then hit them with repeated MMT (remove a few small scale layers) and MT Erosion.
Does it matter that when using the erosion to remove the stars that there ends up being ringing artefacts caused by the star mask itself? It erodes the area around the stars, not just the stars. I can assume that when the stars are eventually put back in that it covers over these areas again. Or does this mean I just have to make a better star mask
Just been having a play, I can remove all but the smallest of stars but there is some residual star mask ringing. This does not to be as big of an issue when removing them but it definitely is when trying to put them back
So, more tweaking and learning in star masking me thinks
Well after some new masks I have made some more progress, there is still some evidence of the star masks in the fainter areas but it almost looks like its a part of the nebulosity anyway. Actually tried adding different coloured stars, need better masks for that Link to version D
I think this is what you had in mind, it wasn't until I did a red desaturation that I realised how much of a red overcast there was. I haven't done much other than remove some of the purple and red.
Not sure whether to start the colour data from scratch again, as the bottom left has started turning grey. There is purple around two of the brightest stars, those I may just end up cloning out when everything else is set.
That's cleaned up the stars a bit. Might be worth trying again from scratch. I usually do that at least a couple of times with most images
Usually when I do a reprocess I start from scratch It has cleaned up the stars but doing so has lost most of the star colour that I was hoping to preserve :/
Usually when I do a reprocess I start from scratch It has cleaned up the stars but doing so has lost most of the star colour that I was hoping to preserve :/
You can add RGB star colours or build star colours from the NB data and add them at the end.
To me it looks really good now Colin - congratulations. Minor disagreement with the colour- seems to be not enough red but it's still good. I usually use RGB for narrow band stars myself but there are probably better ways. I do agree with Mike - a little magenta is fine, just not too much.
You can add RGB star colours or build star colours from the NB data and add them at the end.
Very true, just need to learn to build better star masks, need tighter ones for star replacement.
Quote:
Originally Posted by Stevec35
To me it looks really good now Colin - congratulations. Minor disagreement with the colour- seems to be not enough red but it's still good. I usually use RGB for narrow band stars myself but there are probably better ways. I do agree with Mike - a little magenta is fine, just not too much.
Cheers
Steve
Thanks Steve, I think my colour calibration was a bit off to begin with. Usually the SHO palette ends up being a bit green, looking at it after the red desaturation I think it turned out a bit too red but I do agree, it ended up taking too much of it.
I think the way you do it is the best way, adding RGB stars. I have no idea when we're going to get another clear night down here in Melbourne so I going to do some synthetic colour from the narrowband. I personally find HOS giving some nice star colours but I do need to play with it more.
At the moment my star masks are creating nasty issues when I add them in, another few days of masks make
Very true, just need to learn to build better star masks, need tighter ones for star replacement.
I only modify the colour (CIELAB a* and b*) with a few gentle blend operations and leave the lightness (CIELAB L*) alone. I find that a lot easier than wholesale star replacement.
If you are in PS, don't overlook the selective colour tool - it's powerful. For your original image invoke it, select magenta and then slide the magenta bar all the way to -100%. See the result in the first image below.
I then took liberties and did a couple more adjustments (in the same invocation of selective colour) to produce the second image . Hope you don't mind - I'm learning NB!
PS: Excuse the jpeg artifacts - they're pretty bad, but you get the idea.
PPS: No star mask used because there was no other magenta in the image.
Last edited by marc4darkskies; 20-06-2016 at 10:46 AM.
As Marcus outlined in his post you have to take advantage of the fact there is no purple anywhere else in your picture (or very little) so you can target that color.
Here's a file RAR file here with a stack in PS based on your original picture.
I isolated the two layers of interest, one for the neb and on for the stars. Have a look at the settings. You'll get the gist of things.
In the end it all comes down to proper star masks and how much time you want to spend on it. Mask are easily done but using a difference blend between layers then refine the selection by contracting/expanding then feathering.
Also keep in mind the shadow highlight tool on stars only. It's a flexible way of deringing your stars while controling the cores and colors.
Finally blending layers on top of each other as "color", "lighter color", "Luminosity" helps as well getting rid of hard edges.
I only modify the colour (CIELAB a* and b*) with a few gentle blend operations and leave the lightness (CIELAB L*) alone. I find that a lot easier than wholesale star replacement.
Wasn't something that I'd ever considered, just gave it a quick shot and I think I still need to work on my star masks It is not only changing the star colour back the small area around the stars as well, they're not tight enough so its bleeding through. Or maybe my blending is way to harsh
Quote:
Originally Posted by marc4darkskies
Very nice image Colin.
If you are in PS, don't overlook the selective colour tool - it's powerful. For your original image invoke it, select magenta and then slide the magenta bar all the way to -100%. See the result in the first image below.
I then took liberties and did a couple more adjustments (in the same invocation of selective colour) to produce the second image . Hope you don't mind - I'm learning NB!
PS: Excuse the jpeg artifacts - they're pretty bad, but you get the idea.
PPS: No star mask used because there was no other magenta in the image.
Quite like the result Marcus, I do have PS4 on my computer but I have barely used it for anything other than its Save for Web function
It sounds similar to what I did on version C. All I did was open Version B and do a selective colour desaturation on the stars.
Narrowband is a lot of fun ain't it Marcus, has its own challenges though. As you mentioned on your recent first one, it is less about the colour exactness of LRGB but I personally find narrowband more fiddly than LRGB to process. It would also be fair to say that in mind case I butcher RGB a bit as well
Tonight is just going to be playing with the RGB data, I think I may have found a better star removal technique that will hopefully leave less artefacts.
As Marcus outlined in his post you have to take advantage of the fact there is no purple anywhere else in your picture (or very little) so you can target that color.
Here's a file RAR file here with a stack in PS based on your original picture.
I isolated the two layers of interest, one for the neb and on for the stars. Have a look at the settings. You'll get the gist of things.
In the end it all comes down to proper star masks and how much time you want to spend on it. Mask are easily done but using a difference blend between layers then refine the selection by contracting/expanding then feathering.
Also keep in mind the shadow highlight tool on stars only. It's a flexible way of deringing your stars while controling the cores and colors.
Finally blending layers on top of each other as "color", "lighter color", "Luminosity" helps as well getting rid of hard edges.
I've downloaded it but not yet had a look at it Marc, hopefully find some time tonight
I've done Version G now. Tried adding some more red but I don't yet have the time to try to mask all of those minuscule red stars embedded in the nebulosity. That may be a job for the coming days.
Version F was an attempt to replace the stars using the techinque Rick suggested, using the CIEa&b, works well but there is visible issues due to star masks It wasn't so obvious in the 32-bit format in PI.
Version G was just keeping the original stars but making them look nicer, think I prefer the Hubble stars moreso.
. Hope you don't mind - I'm learning NB!
