DGM NPB filter = psuedo-narrowband with DSLR?

[smokey688i]

I recently decided to mod my Nikon D5100, and try a session on the Flaming Star Nebula and IC405 area using only the Improved DGM NPB filter by Omega Optical. This filter allows through the spectrum lines of Hb (486nm) and both Oiii (495.9nm + 500.7nm) in one passband, and Ha (656.3nm) in a second 10nm-wide passband, with very sharp dropoffs outside of these target ranges. The result image after a typical StarTools process flow (without using the Color module in fact) is actually fairly well color-balanced when using a full spectrum camera.

flaming star area normalprocess small.jpg (479.24 KiB) Viewed 6519 times

But I decided to try to take a narrowband approach to processing it to see if any improved could be gained. I separated the RGB channels of the original linear stack into 3 separate grayscale files, R being quite close to pure Ha, and B+G both representing the Hb/Oiii area. I first processed the R channel to my liking, then I blended the G and B channel files together in the Layer module, and processed it to my liking. Next I created a synthetic luminance by blending the R and G+B files together in the Layer module, giving a significant emphasis to the R file. Then I loaded up this luminance file in the LRGB module as L, the R file as red, and the G+B file as blue, leaving channel interpolation on for StarTools to create the green channel. After some tweaking in the Color module, and a final light denoising in Noiseware, this is my result. I was going for an HST palette appearance, so as to bring out the contrast more of the different areas.

flaming star area small.jpg (467.48 KiB) Viewed 6519 times

I'm liking this result, but I'm curious if anyone has any ideas as to a better route to process this? I've posted the raw stacked fits file here for anyone interested in taking a swing at it:
https://drive.google.com/open?id=0B3nLN ... Hc1VUt1c1E

[admin]

That's a marked improvement in detail - something very hard/impossible to accomplish by taking the data as-is (e.g. all channels weighted equally).
A very interesting post / exercise!
Looking at the data (thanks for uploading that), it appears the red channel indeed contains a great amount of detail, with the other channels containing comparatively little. In particular, inspecting the 3 different channels, strong gradients appear in the green and blue channels, but not in the red.
Another interesting observation is that coloring - when treated as an RGB set - is still surprisingly "interesting", still showing some semblance of cross-spectrum star temperature variations.
Very cool!

[smokey688i]

Thanks for the kind words, Ivo. Regarding the gradients, this shot was taken from my suburban backyard, at an average altitude angle of 45-60deg in the direction of a nearby major city. Despite the sharp cutoffs of the filter outside of the oiii/Hb target range, there is obviously still a lot of light pollution around those wavelengths. But the deep red Ha was very clean indeed.

I was originally hoping for a result that would appear simar to others' narrowband bicolor interpretations, where the contrast between the brighter Ha edges of ic410 showed greater contrast with the center area of that object, but perhaps I need to shoot longer subs in order to pick up more of the faint oiii signal, since by using a single multiband filter I can't emphasize the separate spectrums differently at capture time.

-Chris

[smokey688i]

On another note, what would be the best way to tone down the semi-square reflections around the brightest stars? I noticed these were somewhat less prominent in the green and blue channels...maybe some sort of mask when building the luminance layer? Or would the heal module be better?

[admin]

On another note, what would be the best way to tone down the semi-square reflections around the brightest stars? I noticed these were somewhat less prominent in the green and blue channels...maybe some sort of mask when building the luminance layer? Or would the heal module be better?

They're interesting diffraction patterns (or reflections?) - I'm not sure what would be causing them. Would you have an idea? It's always better to address things during acquisition, rather than during post-processing. There are certainly ways of manually processing them out if you're not squeamish about that sort of thing, including - as you suggest - building a selective mask, using the Heal module, etc.