M42 Processing ideas...

[nightster]

I finally have taken advantage of my Astro Club's dark site obsy the past 2 new moons. I chose M42 as RGB targets are pointless in my LP rich backyard. I gathered RGB a month ago and made the mistake of shooting 360s. My LP brain thought this was appropriate. Of course the core was saturated.  I went back this past weekend and got the L. 10s, 30s, 120so, and 300s.

Now I'm looking for advice to process this. My thought is to process each sub length separate for detail in the various areas. Then to put each image into a different layer and use masks and opacity adjustments to reveal the details throughout the image.

Am I on the right track with this thought? 

[swag72]

This is an excellent way to do it with these high dynamic range targets. http://www.astropix.com/pfa/sample4/sample4.html

You're on the right track with your thinking

[nightster]

Thank you for the Great link, itmade the process even easier than I expected.  I used PI for preprocessing. I also stretched each stack over there. Then saved them as 16bit TIFs. Opened them in PS and followed the astropix tutorials. Except that I used 4 layers instead of 2. This is still WIP and it obviously needs a crop to remove stacking defects.  Even with the issues can you suggest where the image needs more work?  

[swag72]

I think this is looking very good Jeff ........ The core is nicely controlled and the running man is there too.... I think you've got a very pleasing M42

[RayD]

11 hours ago, swag72 said:

This is an excellent way to do it with these high dynamic range targets. http://www.astropix.com/pfa/sample4/sample4.html

You're on the right track with your thinking

What a great link.  I've always struggled with layer masks so this is really useful, thanks for sharing this Sara.

[RayD]

Really nice looking image here, well done and thanks for sharing with us and I think it definitely confirms you're on the right track.

[Uplooker]

Hi Jeff,

That is a mighty fine looking image, well done. Also thank you to Sara for posting that link. As Ray says it is very informative.

[nightster]

I appreciate the feedback. I was able to combine this with the color data from last month. They were not aligned in rotation as much as I would have liked which results in quite a large crop. I think I'll need to be more careful in my rotational alignment next time. Before I post this in the Deep Space forum I'd like opinions if it needs any color adjustments or detail enhancements. 

[AKB]

47 minutes ago, nightster said:

Before I post this in the Deep Space forum I'd like opinions if it needs any color adjustments or detail enhancements. 

Looks nice!  Wonder if you might not want to bring out a bit more of the detail which is all there?  I had a quick go at this...

[nightster]

What process did you use?

[AKB]

3 minutes ago, nightster said:

What process did you use?

Well, I don't use PS or PI, so I can't tell you how to do it there.  I simply created a pseudo-luminance channel by averaging RGB from your posted image, and then applied the processing described here:

Z. Farbman, R. Fattal, D. Lischinski, R. Szeliski, Edge-Preserving Decompositions for Multi-Scale Tone and Detail Manipulation, SIGGRAPH 2008

...then recombined the processed luminance with the original RGB.

Really nice (short) video summary of the process here: http://www.cs.huji.ac.il/~danix/epd/

Matlab code (which is what I used) here: http://www.cs.huji.ac.il/~danix/epd/wlsFilter.m

Not sure if this helps!

AK

[Ruud]

A very good image by itself. 

Here's an extra colourful version (Topaz Adjust 4 - Spicify) and a detail enhanced version (Topaz Adjust 4 - Small Details).

Topaz Adjust is a plugin that works in all image editors that accept photoshop compatible plugins. Very quick and simple to use.

[ollypenrice]

Sara's link goes to an excellent method. It's the one I use, though I met it on Jerry Lodigruss' website. 

I think you used the methods extremely well, Nightster. I'm afraid the two modified versions are not to my taste and show the effects of over-processing using heavy algorithms.

Your own version is a perfectly good starting point although, when layer masking in your short colour subs, I think increasing the colour saturation for the shorts would have been good. The brighter colour signal is, the more easily it 'burns out to white' as has happened to your Trapezium.

What I would now do is stay well clear of the hard drugs and work gently in small steps!

In general try to avoid the colour saturation tool in Ps. It is a noise-increaser. Try these techniques.

1) Save a copy and convert to Lab colour space and increase the contrast in a and b channels by about 35. Re-convert to RGB. If you have simply got more colour, fine. Keep it and continue. If the Labspace image looks black clipped save it and then copy and paste it onto the original, set the blend mode to colour, flatten and save.

2) Make two copy layers. Set the blend mode in the top one to Soft Light and flatten it on onto the middle one. Now set the blend mode to colour and flatten.

Next up I'd work on the local contrasts. Noel Carboni has a great set of astronomy tools (his site is now called Pro Digital, I think) which include Increase Local Contrast. Alternatively you could try Local Histogram Equalization in PI but I would only ever save it and bring it into Ps to use as a layer.

Finally I'd do some shaprening only in key areas using Unsharp Mask in Ps. I do this as a bottom layer and let it through into the top layer only very carefully. I'd be careful never to sharpen stars.

As Sara says, you have a super image so treat it gently and ease out its hidden charms.

Olly

[nightster]

Unfortunately I haven't gotten the short color subs yet. I'll grab some next outing. Should I mirror what I did with the L? 10s, 30s,120s? 

Imaging dark is so much more successful then in LP. I would only get horrible filter independent gradients that ruined many nights of work. I've learned I'd rather drive out to the dark and only get one nights 'good' data vs 5 nights of light pollution soup. 

[ollypenrice]

I wouldn't bother with very short L subs. Just do RGB. Why do we shoot L? To get more signal to noise in less time than we get from RGB. But in this case we are trying to avoid signal!  Also, one of the things about blending in the Trapezium region on this target is keeping up the colour saturation in the bright parts. Lum tends to dilute it. In this rendition the short subs were RGB only.

Olly

 

 

[nightster]

That is some pretty important information that I haven't heard before.  To check my understanding: Use subs of varying exposure lengths for the RGB filters only on these high dynamic range type targets to help block signal in the high signal portions of the target. Use layer masks with those various rgb stacks to reveal the color in the image across the whole frame. Only use the L to quickly gain detail without various exposure lengths. Use Lab color and saturate only the A and B channels to taste. Use Carboni's (PS) or LocalHist (PI) to increase detail in those high signal zones.  Can't you saturate the bright parts in L also? But maybe the detail from the short colors subs will compensate for that.  Also do you always shoot the same time on each RGB channel? 

[ollypenrice]

17 hours ago, nightster said:

That is some pretty important information that I haven't heard before.  To check my understanding: Use subs of varying exposure lengths for the RGB filters only on these high dynamic range type targets to help block signal in the high signal portions of the target. Use layer masks with those various rgb stacks to reveal the color in the image across the whole frame. Only use the L to quickly gain detail without various exposure lengths. Use Lab color and saturate only the A and B channels to taste. Use Carboni's (PS) or LocalHist (PI) to increase detail in those high signal zones.  Can't you saturate the bright parts in L also? But maybe the detail from the short colors subs will compensate for that.  Also do you always shoot the same time on each RGB channel? 

The Lab colour trick involves increasing the contrast (using the bog standard contrast slider) by the same amount in a and b channels. It does increase colour saturation but without the noise of the standard saturation tool.

Olly

[nightster]

I was able to get out last night and I grabbed the short RGB subs for this. 20x10sec ea channel RGB. I used the layer mask and then the convert to Lab color and contrast boost on a and b only. I'm really happy how that little bit of extra data made the whole image pop.

 

[wimvb]

Looks great. I think that you can increase local contrast in small steps, as Olly suggested. In PixInsight you can use either LocalHistogramEqualization or HDRMultiscaleTransform. The latter is a real power grinder, so use it with care. You can use it on a copy of the image, then combine the HDR processed image with the original using PixelMath

F=0.5; F*HDR + ~F*Original

It's easy to change F from 0 (no HDR) to 1 (pure HDR)

BTW, PixInsight also has an HDRComposition process for combining images of different exposures.

[frugal]

If you have PixImnsight then there is the Process -> Image Composition -> HDR Composition. This will take multiple images and automatically HDR them together in a similar way to the Photoshop layers method. 

[nightster]

14 hours ago, wimvb said:

Looks great. I think that you can increase local contrast in small steps, as Olly suggested. In PixInsight you can use either LocalHistogramEqualization or HDRMultiscaleTransform. The latter is a real power grinder, so use it with care. You can use it on a copy of the image, then combine the HDR processed image with the original using PixelMath

F=0.5; F*HDR + ~F*Original

It's easy to change F from 0 (no HDR) to 1 (pure HDR)

BTW, PixInsight also has an HDRComposition process for combining images of different exposures.

I worked this image in steps as I realized I needed data of differing exposures. I processed each stacked separately then did a layer masks in PS to add the new data. Adjusting the layer as needed to blend it into the base layer. I might want to try the run again in PI using the HDR to see how that works. I had a goal of getting more proficient in PS this year. I found that it made my eye a bit better in determining what I thought was needed at each step. I was leaning on PI for 'whole image' algorithms a bit too much and forced myself to look at specific portions of an image. I'm pleased with my progress.