More on the Leo Triplet, or What Happens When You Refuse to Give Up (even longer than my last Leo post)

[rickwayne]

This was really a salvage job. I don't have to pixel-peep very hard to find problems with this image, but given what I had to work with, I'm pretty happy.

I'm posting it here in Getting Started (also cross-posted to Cloudy Nights)  because I had to dig deep in my bag of tricks to get this, and perhaps one of those tricks might be useful. And also to show that a certain stubbornness can sometimes bear fruit. Mind you, starting with lots of nice clean data will do more for you image than any processing tricks ever can.

This is two nights worth on the Leo Triplet, LRGB with a total of three hours integration time, heavily weighted towards luminance. I used quite a low gain on my camera for maximum dynamic range.

First night was endless equipment problems, mostly self-inflicted -- I was running a star party for friends and was pretty distracted. I wound up imaging at a crazy-wrong backfocus from my flattener/reducer, and focusing manually. A bevy of other problems meant that I got the sequence started very, very late, so I only got in 42 minutes and forgot to take flats entirely!

The second night went much better. Just me, and I'd sorted many of the tech problems, but as it turned out, not all of them. I also had clouds roll over, stopping me dead for a good 45 minutes. I had to throw away some of my Blue frames because the last ones featured trees instead of galaxies.

Since the first night had no usable flats, I wound up just stacking the two nights separately and then stacking each channel's integrations. I might get less noisy results if I went back, started from scratch, and did them as two sessions in Astro Pixel processor. But due to the backfocus and other problems, I knew there was only so much I could wring out of these data anyway.

Once I had done the preliminary processing of the individual channels, I had a lot of trouble getting a decent color balance in Astro Pixel Processor with either night's data. (Still working with them to understand why.) So, rather than struggle with its somewhat mysterious sliders, I used its terrific light-pollution reduction and stretching tools, exported TIFFs, and worked with them thenceforth. The LPR tool allowed me to cheat somewhat and fix some of the problems that the missing flats had caused.

Opening all four channels in Photoshop, from Red, I did a simple Select All and copy. Then File | New to create a new image, changed the mode from grayscale to 32-bit RGB, and went to the Channels tab in the layers dialog. Pasted the copied data into the Red channel. Repeat for G and B, and some nice galaxy color appeared.

Next, to exploit the Luminance layer, I converted the image to Lab Color, and copy/pasted the luminance in as the Lightness channel. Simple as that -- all of a sudden,  I now had a nice deep tonal range of plenty of details in the galaxies.

(You can also use this trick to create a synthetic Luminance image for contrast enhancement or other tweaks -- just go to Image, Mode, and  select Lab Color. Changing back and forth from Lab to RGB is nondestructive.)

Before I touched anything else, I exported to a TIFF and ran Topaz Denoise AI on that. Topaz recommends that you run Denoise as early as possible in processing. It did a very good job muting the griblies without giving anything a plastic feel or knocking out detail.

Now the image needed some enhancement work on the galaxies, and also some saturation to bring out star color. For an image like this with few, distinct areas of nebulosity with little overlap with stars, masking would work fine. But I chose to go with star removal for a bit more control. So, I pulled the denoised image back into Photoshop, only to turn around and export an 8-bit TIFF for starnet++ to play with. Pulled the resulting starless layer into Photoshop yet one more time, set it back to 16 bit RGB, and set its blend mode to Difference, producing a stars-only image. Stamp Visible to turn that into its own, stars-only layer, which I then set aside by making it invisible.

Turning the starless layer's blend mode back to Normal, I did some pixel editing with the Spot Healing Brush to remove some of the "star ghosts" that starnet can leave behind. A little bit of contrast enhancement; really, APP's stretch did all the work, I just tweaked a teeny bit. Then yet another TIFF export, this time for Topaz Sharpen AI. Usually I find the Motion Blur mode works best, but tonight the Unsharp mode  did a better job. I pulled the slider WAY over, and some really nice detail was extracted in the galaxies, especially in M66. Pull the result...you guessed it...back into Photoshop.

Now, the stars layer. Add Saturation layer (option-click on a Mac, think it's alt-click on Windows, to assign its effects to the stars layer only). Pull the slider way over. Look at all them pretty colors! Whoa, look at all them ugly halation artifacts! (Told you there were some deep-rooted problems with these data.) Maybe not quite so much saturation...and out comes the Spot Healing Brush again, to clean up some of the real eyeball-pokers. Color balance layer (same option-click trick) to make the star colors a little more reasonable. (If I were going for accuracy instead of art, I'd use the star-color calibration tools in APP, or in PixInsight if I had that.) Add a Levels layer too to black out the background so it doesn't have a color cast.

Lastly, set the blending mode of the stars layer to Screen, so that it combines additively with the starless layer below it.

 

Hope this is helpful to someone. Not everybody, not even everybody who has Photoshop, owns Denoise and Sharpen AI, of course. Still, they're good investments, and of course the principles apply to other tools as well.

[ollypenrice]

OK, no replies so far so I'll jump, provided you don't mind my being honest. If you do mind, stop reading now! 

This seems like an awfully complex process for a result which I don't think has hit the data's sweet spot, though I haven't seen the raw data so I can't be sure of this. However, I've been doing this for a long time and have a vague notion of what may lurk within.

To my eye there is very little galaxy colour showing yet I would be surprised if it were not present in the data. (I could be quite wrong.) The trick is to find that colour without provoking an orgy of colour noise in the rest of the image but Photoshop has all you need for that. Similarly Photoshop has, I think, all the noise reduction and sharpening that you need provided you apply it selectively.

There are two big patches of dodgy background sky, shaped like footprints, one around the brightest star and one below it in the middle of the trio. I'd want to know how those got there. They should not be hard to fix.

But my main reaction is to feel that, the simpler the processing, the better.

Olly

[alacant]

On 16/05/2021 at 04:37, rickwayne said:

the principles apply to other tools

Hi

Lovely story and nice image. You're starting to get the faint tail too I think. Another session should do it.

I had a go at the stars and the colour, but I'm far from sure...

Cheers

Edited May 18, 2021 by alacant

[rickwayne]

WRT color in the galaxies, that's another thing I apparently overlooked -- my initial go at these data definitely had more. So that's something to investigate. The lovely thing about non-destructive layers in Photoshop is that the question "where did I go wrong?" actually has an answer!