TVGDenoise and mask

[whitenerj]

Just a basic question. Do you guys use a mask for the non-linear TVGDenoise process?

Jason

[jerryyyyy]

I do for cleaning up the background.  I use a L mask set to 0.25 shadow... you want the background black in the mask so that the background is what gets clobbered. 

You can also use the residual mask from the ExtractWaveletsLayers script.  I make a small preview and look at what happens.  I also try ATM at various levels... have found that about 60% of the time I prefer ATM.... may be smoother without mottling but lose detail.  How much time do you have to try variants.....

I am a relative novice here.  maybe some of the gurus will chime in about what are the nuances of cleaning up and differences between masks. 

You have to watch the TGV so as to not be too aggressive and get mottling. 

[whitenerj]

See that's the thing, mottling seems to sneak up on me sometimes. When you use TVGDenoise does the mottling pop out right away so that you know you have over denoised the image. What are the signs you have over TGV'd the image?

When using say a luminence mask and protecting the high signal areas, is it a good strategy to kill two birds with one stone by hitting the background hard and the high signal area softer at the same time?

Jason

[jerryyyyy]

See that's the thing, mottling seems to sneak up on me sometimes. When you use TVGDenoise does the mottling pop out right away so that you know you have over denoised the image. What are the signs you have over TGV'd the image?

When using say a luminence mask and protecting the high signal areas, is it a good strategy to kill two birds with one stone by hitting the background hard and the high signal area softer at the same time?

Jason

Not the most experienced person, but I would make a few preview windows over high and low signal areas to see what you get.  I have process icons set to different settings so i can try the out quickly. 

Not sure you can do both at once.  I denoise background first after going nonlinear then enhance the light areas.  Could be argued to do in reverse order so you do not lose any signal. 

[twade]

Jason,

Previews are your friend here.  Granted because of the nature of TVGDenoise, you won't get a perfect match between your previews and final image, but they will be close enough.  I generally create 2-4 previews and clone each one.  I select 1-2 previews in areas I don't want to loose detail, and I select 1-2 previews in the background regions, especially where there is subtle contrast between the background and dark nebular regions.  I close each preview so I can easily see the before and after.  On my DSLR images I get good results with the following

Strength  0.015  -2
Edge Protection  0.02  -2

Start with these and see what happens.  If you see the "waxy look", decrease the strength. 

I hope this helps,

Wade

[MortenBalling]

I use masks for absolutely everything.

When using an inverted luminance mask for TGVDenoising, the brighter noisy pixels are more masked than the darker ones, which isn't optimal. Therefore I use a slight convolution (blur in everyday language) on the mask first.

I also use TGVDenoise on my flat stacks. That gives a measurable increase in SNR.

Morten

[jerryyyyy]

I use masks for absolutely everything.

When using an inverted luminance mask for TGVDenoising, the brighter noisy pixels are more masked than the darker ones, which isn't optimal. Therefore I use a slight convolution (blur in everyday language) on the mask first.

I also use TGVDenoise on my flat stacks. That gives a measurable increase in SNR.

Morten

Well, you are absolutely right on the TGV Flats.  I just ran some 30 H-alphas on linear and got for noise these numbers:

Regular Flat: 8.369 -005e
TGVed   Flat: 8.213 -005e

You know for a mask you might consider using the extract wavelets residual image.