That blood super moon was fun wasn't it?



I don't have a lens longer than my Pentax D-FA 100mm 2.8 Macro, so I had fun with my new photoshop cc subscription.

For my own recollection later, and If anyone cares about the process I used:
1) Tale your 16 consecutive shots each 15 seconds apart @ F/4, 1sec, ISO1000. Do the same with the lens cap on.
2) In photoshop, pull in the 16 "dark" shots with the lens cap as layers, create a smart object out of all the layers, then set the stack mode to median. This gives a nice dark bias frame to get rid of sensor bias.
3) Then pull in the 16 "light" shots without the cap as layers, and do the same smart object->median trick. Since the stars move, they won't appear in the median. The moon will though, so duplicate the layer and set the stack mode to minimum to get rid of it.
4) Then combine the two layers using a layer mask - the median one is preferred, with the minimum one only over the moon area.
5) Then copy the dark bias frame and merged it into this frame using the "subtract" blending mode in order to remove the bias.
6) Then Blur->Gaussian 100px. Heal the spot over where the moon was to make it smooth. Now we have an atmospherics frame with all the reflected light from the city.
7) Then take the atmospherics frame and dark bias frame and "added" them using the linear dodge blending mode, and merged into a single "offset" frame.
8) Now make 16 copies of the offset frame and "subtract-merge" them into each of the 16 "light" shots - we now have pseudo-calibrated images (flat field seems negligible on the 100mm macro at f/4)
9) Crop out a section of the image that contains the moon across all layers, then save it as a different PSD.
10) Follow the guide here to do super resolution: A Practical Guide to Creating Superresolution Photos with Photoshop
10a) Some notes: auto-align wouldn't work, i had to register them all into place manually (nudging helps). I also found the smart sharpen gaussian mode better suited than lens blur as in the guide, followed by another small amount of unsharp mask.
11) Cool now we have our moon - select it with the magic wand until you get the whole thing. Then expand the selection 5 pixels or so, invert selection, feather selection 2px, and hit delete to clear out everything but your moon.
12) Back in the original PSD, it's time to align the stars - again auto-align doesn't work. I found aligning in the center, then distortion-type transform from two corners seemed to work (the star movement isn't linear across the scene). Don't worry about the moon not lining up, it doesn't move the same speed as the stars.
13) Once you've arduously registered the star fields, convert to smart object->median. Rasterize and heal out the moon. Now you've got a star field worth enhancing.
14) Add adjustment layers with curves, exposure, hue/saturation, and channel mixer until you get your stars where you want them without much noise (blue channel had the most noise for me).
15) Copy and paste your super-resolved super moon onto the spot where the you healed out the other moon (yes it'll be twice as big, but that's OK right?)
16) Export as TIFF, then bring it into lightroom for any final touches.

Yeah that was probably too complicated for me to do for this 100mm shot, but it was fun to learn how to do it.

The single frame starfield, single frame moon, and post super-res moon are posted below too so you can see the difference it made.

Do any of you have suggestions on what I could have done better? I'm fairly new to astro imaging. If I had done it again, I would have taken 32 shots instead of 16. Then I wouldn't have had to worry about the extra minimum stack mode frame due to slow moon movement. I also do wish photoshop had an auto-align that worked on starfields - is there a good plugin to do such a thing?

[ckoehler]

Very cool! Good job! Thanks for the step by step!

[jeffshaddix]

Thanks ck. I'd love to try this someday with a longer lens.

[mole]

Indeed excellent results from your patient, painstaking labor. Thanks for sharing both the image and the method behind it!

[jeffshaddix]

Thanks mole, it was great fun.

I want to try that super-res technique on some other subjects now; it's basically what the K-3ii does internally, so I bet if you could extract the 4 raw images you could use this technique to good effect.

[GeoJerry]

A very impressive workflow. How did you (or anyone) ever figure this out? It seems impossibly arcane to me, but then again I've never used Photoshop. I guess I need to get on the ball there.
It came out really well, especially considering 100 mm!

[jeffshaddix]

Thanks Jerry! I've got a background in image processing (Elec. Engr. degrees), so the technique was fairly straightforward. The image is composed of sensor counts from starts/moon, atmospheric reflections, and sensor bias, the goal is to remove the latter two, which can be done through the image stacks. So I had a theoretical idea of what needed to be done, and thankfully photoshop hasn't changed much in a decade, so the mechanics side was pretty simple. The super-res article was a neat find though (through a post on these forums I think).

Cheers!

---

Last edited by jeffshaddix; 10-14-2015 at 07:30 PM.

[Ex Finn.]

Very nice result from your PP!

[GeoJerry]

Thanks Jerry! I've got a background in image processing (Elec. Engr. degrees), so the technique was fairly straightforward. The image is composed of sensor counts from starts/moon, atmospheric reflections, and sensor bias, the goal is to remove the latter two, which can be done through the image stacks. So I had a theoretical idea of what needed to be done, and thankfully photoshop hasn't changed much in a decade, so the mechanics side was pretty simple. The super-res article was a neat find though (through a post on these forums I think).

Cheers!

This may be asking too much of you, but if you have time could you explain what you did in layman's terms, maybe summarizing the basic concepts?