Elephant Trunk, RGB + SHO at 2800mm

I think it looks really good, and that's an understatement.  The larger stars are potentially bloated, but it's at 2800mm so if course they are.  I just think it looks great.

As an SCT guy, this has long been an elusive target and you have done it more than justice.  I thoroughly enjoyed it.

But don't let it bother you.  It's just the nature of the optics.  The contrast you achieved in the nebula far outweighs the mild star bloat, and in a way it demonstrates the dynamic range between the dim nebula and the very bright stars.  Very enjoyable astrophoto.

@Sven Arnold, thank you.

You know, I must have run SCNR about a dozen times with different values (which I think has a  similar affect for green) and settled on 0.6, which did leave some in, but then… well, it rather vanished as I did further balancing of the background.  Your point is well taken, I do not have some religious objection to green.  Saturation – I'll give it a try next pass through this data, was it any particular color you though needed more saturation or just overall, I added quite a bit to yellow trying to get it more gold.

Looks a super image and also very well framed.  The only thing that I would change is have the stars less stretched and smaller before adding them back into the starless image- so that the two stars of the double in the elephant's trunk are separate.  

Tim

(Oh and as a complete btw Linwood I  wanted to thank you for a detailed thread that you contributed some time ago - I think it was on CN -  in which you posted an analysis of some RA oscillation problems with a CEM70 mount - Reading it was very helpful in helping me to resolve some similar problems recently)

Hi Linwood.

See what you mean!  Don't get why an SCT should be inherently worse than a Newtonian wrt star bloating and removal?  --  had imagined that the Newt  would be more inherently challenging because of the  diffraction spikes?   

Can only rather lamely ask whether you on the latest version?  For me StarExtractor ver 2.03  did a clean star removal (applied to the non-linear NB) image) of IC1396 (~ 9h HA + 6h  Newtonian OOH image at F 4.0)  even with tiles set at only 20%.  https://www.astrobin.com/kqyufn/0/#rC

Tim

Also wondered whether star removal prior to addition into the full NB starless image might work  i.e. wondering whether just one component in particular has the bloated stars ?

Try as I might, I couldn't tame your stars, but here is my best attempt.  All I can suggest is do a very small linear stretch before removing stars with StarXterminator with unscreen selected.  You can then work on the stars and starless separately and then do a re-screen later.  Your starless stuff looks stunning:

OK, so downsampling 4X and using starnet and the adam block star reduction twice I ended up with this.  I'm sure that you will end up with a much better image with downsampling 4X and using StarXterminator; I'm looking forwards to seeing it as I think it will be stunning.

OK, I worked through this again, with some kind help from Russ at StarXTerminator, and got a better handle on the stars. 

I also included more green and color. 

Thank you all for suggestions, the new version is: 

https://www.astrobin.com/85ed5q/B/

I'm not delighted with the halos around some of the stars, but if you don't look too closely you will not see them.  I'm not quite sure how to get rid of them with the reduced stars. 

Again, would welcome criticism and suggestions.

Linwood

In this case I maintained the heavy over-sampling of the main image.  That may be pointless but I wanted to work through the process required to do so. I mention it as it is relevant to your point.

To produce a starless image that is very over-sampled, but without reducing it, I found you could do this: 

1) After stretch Clone it and down-sample, say 0.33x is what I used
2) StarXTerminator on that withOUT the unscreened setting
3) Upsample to the original (exact) size
4) Subtract those stars from the original

This yields a clean image, but it does not do so if you use the unscreened feature.  This also means the stars I later added back in had to be replaced with a simple addition rather than the unscreened method you have above. 

This was all probably a pointless exercise as the 0.18" scale was way over-sampled, and I could have downsampled before starting.  The problem was I had to go at least 3x down.  2x down didn't remove the stars.  And 3x would put me around 0.56" or so, which seemed a bit too small for effective deconvolution (though I am not sure that is true). 

If I do it again, I plan to downsample 3x and do it more conventionally with the stars, the unscreened technique works nicely I think and would improve the scars left behind.

The color was weird, I used: 

        R:  (1*Sii+ 3*Red) / 4
        G:  (1*Ha + 3*Green) / 4
        B:  (1*Oiii + 3*Blue ) / 4

I played with more and more narrow band, since I had so much of it, but the colors really improved with the BB, for reasons still a bit mysterious to me (all these were LF first to Ha). 

I then arcsinh stretched this to the max, then histogram stretched it so that was dim but not bad.  I extracted these stars (before Luminance combine) 

The synthetic luminance was mean(red, green, blue, Ha, Sii, Oiii).  I played with a number of variations on that, but it was pretty good - histogram stretched, and curved to give contrast I liked. 

I deconvolved the synthetic luminance pretty heavily, then ran EZ Denoise on it.

I then LRGB combined that but adjusted the channel weights shifting away from Green: .9, .8, 1.0, and increased saturation.

What's bizarre about all this was I had so little RGB data, but it seemed to provide a lot of the texture, this despite if I did a pure RGB combine I thought it looked too flat, without much detail. 

I added a bit of SCNR later at 0.5 also, didn't want too too much green.

Star De-emphasis was run on the final result after all these steps. 

And to your question: stretch to taste is generally what I did, including the curves.