My first attempt at processing M31 received some welcome responses and useful feedback (mostly from a more specialised forum) but I felt there was more to be had from the data so I went back to the original stacks and reprocessed. More about the changes but let's cut to the chase:
M31 - The Andromeda Galaxy
'Scope: TEC 140
Mount: ASA DDM60 Pro mount
Camera: FLI ML16803
Subs: 5 x 1,000 seconds using each of Astrodon Tru-Balance RGB filters plus 10 x 1,500 seconds using an Astrodon 3nm Ha filter
Software: PixInsight and Photoshop CS6
I'm also happy enough with the result to offer the image for inspection at its full resolution of 4096 x 4096 pixels - click here
. For comparison the end result of my previous sequence of attempts at processing the image can be seen here
Previously I had done an HDRMultiscaleTransform in PixInsight on a luminance image (after masking the stars) and used that result as my starting point. This time I exported an unprocessed image, after the usual stretch, into Photoshop where the fun began. First step was to manually remove all the stars using Photoshop's Spot Healing brush. In its context-aware mode (CS6) it does an absolutely remarkable job, rarely needing to be undone, so long as the brush size is tuned to the star size. As an aside, if anyone manages to automate that tool (select all stars in turn, choose an appropriate brush size and click) they would deserve a knighthood!
That left me with an exceptionally clean "No Stars" version which I then used to generate a luminance "Just Stars" image (Photoshop layer, Difference mode). The "No Stars" image went back into PixInsight where I generated two new versions, one with the aforementioned HDRMultiscaleTransform tool and the other with the LocalHistogramEqualisation tool. Those images were blended together in Photoshop with the HDRMultiscaleTransform only being used to provide a slight boost in contrast. The "Just Stars" luminance image also passed back through PI where I applied some Deconvolution to reduce the star sizes.
As previously described the Ha processing combined the 10 x 1,500 second subs to produce an image with a fairly high Ha contribution from the stars as well as from the emission nebulosity. Lacking a narrowband continuum filter I used PixInsight's PixelMath to subtract a portion (1/15th) of the Green signal from the Ha one using linear (unstretched) data. Green because my assumption is that that colour is pretty isolated from Ha and Hβ emission, the proportion being chosen by trial and error to remove as much as possible of the Ha signal while not clipping any areas "below" black. That left me with a very much reduced core which was actually pretty distinctly separated from the spiral distribution of the emission nebulae and this time I used a mask and a curve to selectively de-emphasise the Ha contribution from the core even further. I still had to remove quite a number of stars, carefully avoiding removing small patches of nebulosity by doing a blink comparison with a "Stars Only" image derived from the RGB data but once that was done the image was tidied up and coloured so that it could be added to the final image stack.
In addition to the original luminance image I used PI to produce a colour boosted, and hence noisier, RGB image. That went through a much more automated star removal procedure in Photoshop which, after minimal manual clean up, yielded "Just Stars" and "No Stars" colour images. The "Just Stars" image needed a little extra cleaning where colour had got burned out in the cores and star colour was boosted further before that image was used to provide colour to the deconvoluted "Just Stars" luminance image. Similarly, the "No Stars" colour image received some heavy duty noise reduction and selective colour boosting before being used to colour the luminance "No Stars" image.
If you are still with me you may be wondering why I made it so complicated? Well, it might be ineptitude on my part but I think
this methodology gave me the finest degree of control over the contributions of the various components that were brought together to produce the final result. I've gone for more bling
in this version of the image as it was produce with a view to printing later on. But quite apart from the colour changes I think I have much more successfully brought out the detail while avoiding some of the processing artefacts I managed to introduce last time around. I hope you agree.