Two versions of Andromeda. Any advice?

[Felias]

I have been working on processing M31 again (original post here), following some of the advice I have been given in the forums. This time I have used Sequator instead of DSS. I have also binned the photo, applied HLVG filter, and stretched in PS trying to keep better colours. I have tried not to clip the background excessively, the colour sampler in PS detects RGB values around 20 in the darkest places. I have uploaded both versions below; any further advice will be most welcome!

Old version:

 

New processing:

 

[vlaiv]

Try not to stretch multiple times to avoid "posterization" in M110 and M32. There seems to be three distinct "rings" instead of smooth galaxy:

There seems to be gradient present in your image - try to remove it. Also, try to get background neutral. In second version it has red appearance. Try to make it neutral gray.

Denoising is your friend when you use it selectively.

[tomato]

I prefer the colours of the galaxy on the second version, but I agree with Vlaiv’s observations on the background and the smoothness of the satellite galaxies.

Image processing, don’t you just love it?😉

[Felias]

Suggestions noted; I will try those, thank you all. I actually noticed that the satellites were worse off this time, but at that point I didn't feel like going back. I should have known better. ☺️

Processing is so fun, and it has the advantage that I can do it with one hand while my three-year old daughter immobilises half my body while leaning on me on the couch!

[Ags]

1 hour ago, vlaiv said:

Try not to stretch multiple times to avoid "posterization" in M110 and M32.

Why would this cause posterization? 

[Felias]

1 hour ago, vlaiv said:

Try not to stretch multiple times to avoid "posterization" in M110 and M32. There seems to be three distinct "rings" instead of smooth galaxy:

There seems to be gradient present in your image - try to remove it. Also, try to get background neutral. In second version it has red appearance. Try to make it neutral gray.

Denoising is your friend when you use it selectively.

The red background, I can do that easily, so here:

Would you denoise the nebulosity and leave the background as it is? What about the dust lanes? 🤔

[vlaiv]

2 minutes ago, Ags said:

Why would this cause posterization? 

Start with this:

and do this:

See what happens?

Multiple rounds of curves / levels can produce "cumulative curve" that looks like one I produced and that creates posterization.

[vlaiv]

5 minutes ago, Felias said:

Would you denoise the nebulosity and leave the background as it is? What about the dust lanes?

In general - noise in astronomical images is related to presence of signal. Strong signal usually is noise free - like galaxy core and surrounding parts. Weak signal like spiral arms and outer galaxy halo and background that is mostly signal free is source of highest noise - or rather noise that is most easily seen as there is no signal to "mask" it.

For that reason, I advocate masked denoising. It is fairly easily done. In Gimp - create layer copy of your image and apply denoising on it. Then add layer mask and use image itself (intensity of pixels) as mask (inverted).

This means - where signal is weak or image is dark - it will use denoised version, but where signal is strong it will use original version that is not blurred (denoising causes blur).

Here is simple example of this principle in action (although it would work better on 32bit version instead on 8bit jpeg):

[Felias]

1 hour ago, vlaiv said:

In general - noise in astronomical images is related to presence of signal. Strong signal usually is noise free - like galaxy core and surrounding parts. Weak signal like spiral arms and outer galaxy halo and background that is mostly signal free is source of highest noise - or rather noise that is most easily seen as there is no signal to "mask" it.

For that reason, I advocate masked denoising. It is fairly easily done. In Gimp - create layer copy of your image and apply denoising on it. Then add layer mask and use image itself (intensity of pixels) as mask (inverted).

This means - where signal is weak or image is dark - it will use denoised version, but where signal is strong it will use original version that is not blurred (denoising causes blur).

Here is simple example of this principle in action (although it would work better on 32bit version instead on 8bit jpeg):

Nice, thank you! 👍

[Felias]

I have reprocessed M31 from scratch, hopefully this version is an improvement? I have only stretched it once using arcsinh function, and I have denoised it selectively. It is less contrasty than the previous versions, so the dust lanes are not as defined, but the nebulosity looks better I think. I have pushed the colours a bit more, to compensate from the loss in contrast. Any thoughts?

 

[tompato]

Just want to say this is a lovely image, if I can produce something half as good one day I'll be a very happy man.