|
OK, so here is my effort. Its so-so and I’m sure it can be improved.
I am more impressed with the natural colours of Rick’s version though the brownish areas in the spiral of Richard’s image are also very appealing.
I’m keen to hear how others have approached this.
Here was my workstream – apply at your peril!
DBE on R, G, B and Lum
RGB
Linear Fit applied to R and B using G as the base
Combine RGB (using LRGB Combination)
DBE on the combined RGB
Background neutralisation using a preview from a fairly neutral area – hard to find as there is blotchy green and red favoured areas
Colour calibration using the same preview for background and a preview over the galaxy for white – structure detection off
Copy RGB
Histogram Transformation on original RGB
Channel Extraction – Lum to produce a Lum mask.
Copy Lum Mask and apply HDR Multiscale Transform – 4 layers with 3 iterations to remove the galaxy to make a Star Mask Base
Star Mask on Star Mask Base - Noise 3.5, Scale 6, Structure growth 1, 0, 1 and truncation 0.67
Masked Stretch on copied RGB
Applied star mask to Stretched RGB and added mask stretched RGB to RGB with Pixel Math ie replaced stars of stretched RGB with stars from masked stretched RGB
Built a galaxy mask with PixelMath using Rick’s funky “iif(star_mask1>0.1,0,$T)” expression to apply the star mask to the Lum mask and then made it fuzzy with Convolution
Reduced green with SCNR – don’t think I used a mask
Applied galaxy mask to RGB and boosted saturation with Curves Transformation – hit it a few times to bring out the colour – I am a sucker for bright colours in a galaxy
Adjusted black point with Histogram Transformation
Lum
Built lum mask by stretching a copy of Lum with HT
Produced PSF with Dynamic PSF – about 80 stars
Star Mask of Lum – using same technique discussed above
Deconvolution – 25 iterations, deranging using star mask for local support set to 0.7, global dark at 0.01. All other settings at default
Tried to reduce noise in background with MLT – couldn’t produce a linear mask that masked enough of the galaxy so gave up on linear noise reduction
Stretched with HT
Adjusted black point with HT
Produced a Lum mask and then used PixelMath to generate a galaxy mask
Reduced star sizes with Morphological Transformation with star mask applied – Erosion, size 5, amount 0.35 with 3 iterations
Sharpened stars with Multiscale Median Transformation with star mask applied
Sharpened galaxy with MMT with galaxy mask applied – more aggressive than star sharpening
Noise reduction with MLT with galaxy mask applied
Adjusted black point with HT
Increased contrast in galaxy with Local Histogram Equalisation with galaxy mask applied – amount 0.68
LRGB
Added Lum to RGB – Lightness 0.48, saturation 0.42 and Chrominance Noise Reduction on
Boosted saturation with Curves – galaxy mask applied
There are so many things in this image that I wrestled with – unsuccessfully. Here are a few:
1. I wasn’t able to get the noise in the background down. The more I pushed it, the more detail I lost in the outer regions of the spiral arms. I have 139 subs in the Lum and still can’t quite get a sharp enough distinction in the fainter areas.
2. The stars are not nearly as crisp as I’d like – the outer halos tend to get smudgy. I do see however that this is apparent in the other images as well.
3. Colour balance is challenging and I tend to have too much blue and red and not the warm hues of the others. Would love to now how you guys managed that.
4. My star masks aren’t quite the right size – so when I boost the saturation, the stars are mostly but not completely masked. So there are some nasty green and blue outer sections in the larger stars.
5. I’d love to know how to isolate the smaller stars. Giving a saturation boost to the larger stars leads to many deep yellow/red stars in the background. Is there a funky way to mask all but the largest stars?
One thing that I was keen to see – whether I am being skimpy with my imaging. On a galaxy I am for about 6 hours of Lum and 4 hours each of RGB – all unbinned. Looking at what R & R could pull out of these data, I suspect I need to spend more time on the imaging – or look at larger targets.
Comments/criticisms most welcome
Pete
|