Opinions on sharpness vs noise

[Datalord]

I'm having another one of those painful moments where I must trade off sharpness for noise and I can't decide which way I want to go. So I go to the audience, because it is never wrong. The picture is not fully processed with curves etc, but close. This is with and without an MLT in PI.

A or B?

A:

B:

[Datalord]

kept on trucking and put it up here: 

 

 

[Adam J]

On 05/05/2019 at 22:23, Datalord said:

I'm having another one of those painful moments where I must trade off sharpness for noise and I can't decide which way I want to go. So I go to the audience, because it is never wrong. The picture is not fully processed with curves etc, but close. This is with and without an MLT in PI.

A or B?

 

I would go for A, the soft stars in B are making my eye go funny. You could try to isolate the stars when applying noise reduction to the background.

Adam

[Datalord]

1 hour ago, Adam J said:

I would go for A, the soft stars in B are making my eye go funny. You could try to isolate the stars when applying noise reduction to the background.

Adam

Thanks Adam. It's a week since I did my pixel pushing on this one and looking at it now, I can barely see the noise difference, but I definitely can see the sharpness difference. The plight of zooming in way too far while processing.

[iansmith]

I prefer the smoother finish to the galaxies in B, the softer stars doesn’t bother me. But as Adam suggests, mask them off before applying Noise reduction or mask the rest off and use MT to tighten them up after noose reduction. 

Cheers, Ian

[ollypenrice]

Do you need to trade off sharpness for noise at all in this image? I can't see why you should have to do so. I don't see any galactic details that really beg for sharpening and the noise is, as ever, confined to the fainter signal which can be isolated via masking for NR. If you want to tighten up the stars I'd do that by isolating them first. 

I don't use PI for this stage of processing but the principle of isolating the different parts of the image for either sharpening or NR is the same whichever graphics program you choose.

In Photoshop it would be child's play to put one of your versions over the other and then select and erase those parts of upper one which you dislike.

Olly

[Datalord]

On 20/05/2019 at 16:44, ollypenrice said:

Do you need to trade off sharpness for noise at all in this image?

Nope, that's my takeaway as well. It's just hard to see the wood for trees while I'm deep in the process.

On 20/05/2019 at 16:44, ollypenrice said:

In Photoshop it would be child's play to put one of your versions over the other and then select and erase those parts of upper one which you dislike.

Can't believe I haven't thought of this before. I do it all the time as a final step on selective sharpening, but I haven't thought of taking entire version into PS like that.

[ollypenrice]

3 hours ago, Datalord said:

Nope, that's my takeaway as well. It's just hard to see the wood for trees while I'm deep in the process.

Can't believe I haven't thought of this before. I do it all the time as a final step on selective sharpening, but I haven't thought of taking entire version into PS like that.

I use a few PI routines along the way but rather than fight with PI masking I just import them into Ps and use layers, the selection tools and the eraser. I like to see what I'm doing in real time.

Olly

[Datalord]

Fully agree, the masking is definitely a weakness in PI. I saw some tutorial where they reluctantly mentioned you can clone stamp in a mask, but it feels like they're being a bit too idealistic about the whole thing.