TGVDenoise 1.0 Released

[Carlos Milovic]

You may also try increasing the number of iterations. I think that 100 is quite a low number. 500 is much closer to the real convergence point of the algorithm. This may change in future releases, with multiscale aproaches. But, right now, try more iterations.

[pfile]

 8) okay. i had increased to 300 but i guess that was not enough. i'll keep fiddling.

[pfile]

is there any advantage to working on a linear image with this process? the image in question has already been autostretched by lightroom and exported as tiff.

[Carlos Milovic]

I don't know

Mathematically, this algorithm (and most other ones) assumes constant gaussian noise. Linear astronomical images suffer from poisson noise mainly. So, in that case is mandatory to use a local support to help the algorithm to discriminate between different SNR zones. In stretched images something similar happens, but now the noise distribution is not as easy described. Also, noise amplitudes in the shadows are larger than on the highlights. So, fine tunning the parameters may be trickier, to protect real details in high snr areas. Again, the local support can make the difference, since it acts as a modulator of the strength parameter (inversally proportional, and in aspect similar to an snr map).

The bottom line. Using LS, both approaches should work. It remains to be seeing wich ones proves to be more user friendly, with easier parameters to fine tune. The next algorithm that is under production should create internally this support maps, and require only a first guess od the strenght that is not critical for the result. There is also an open development field, regarding spatially variable noise amplitudes. We'll get into that once all the current development codes are implemented as official releases, and there is still left some work to do.

[chris.bailey]

Only given it a quick trial but does seem to do better than ACDNR in avoiding the dreaded blotchy background.

[Jules]

Juan, Carlos

Thanks very much for the time and effort, just when I thought I was getting use to ACDNR

Julian

[MickyWay]

Another super tool from the PI team.

It has all the sliders needed to reduce the dreaded noise and I find the default settings are a good starting point.

What other software offers improvements and support we continually enjoy with PI ? And at no extra cost !!

Thanks to all.....

Regards, Colin

[AstroScience]

This is just getting better and better! Thank you for all the hard work.

[northern_nights]

Guys:  I always appreciate your hard work at improving your software.  The sophistication and capability of PixInsight is simply remarkable.  You can made a blink person see the light (figuratively speaking of course).  Many years ago I would never have imagined that I could produce the astro images that were only attainable at world class observatories.

A truly loyal user of PI.

[AstroScience]

Strange, I saw how PI downloaded the update, but after restart I can't find it anywhere.
Any ideas what went wrong? I'm on 1.8 RC7 Win 7 x64.

[Warhen]

Hi guys, Thanks from me too for all your hard work in the past. I noticed the term 'diffusion' in Juan's post. This isn't the 'anisotropic diffusion' replacement for ACDNR that's been hinted at in the past is it?

I'd like to see the NR tools streamlined at some point. I'm a minimilist by nature and the choices seem overwhelming already. As we all know, one can spend an hour fine tuning any one process. Give me two great ones, rather than ATWT, MMT, ACDNR, GREYC, TGVDN, Anisotropic Diffusion...
That's just me.  FWIW, I find LRGBC's CNR very powerful!

Carlos can you compare TGVDN w/ say GREYC from a practicle standpoint- talk of algorithm's will be wasted on me. ;>)

Do you see TGV as a fine-finisher or as a primary denoiser?

Lastly, I find ACDNR's specific approach for chrominance noise to be something important for NR processes. I hope its replacement addresses this.

Thank you and keep up the work!

[Carlos Milovic]

Hi Warren. Yes, this is the replacement we talked about earlier. Indeed TGV may be interpreted as anisotropic diffusion, if we see the iterations as a time evolution. The main difference from TGV to other anisotropic diffusions (normal total variation, and graycstoration too, up to some degree) is that they assume that images are patches, of homogeneus pieces. In other words, images are piecewise constant. TGV, on the other hand, assumes that images are piecewise smooth. This is the advantage of TGV, and why it generates far less artifacts (like, staircaising).

I'm not an expert on the interns of GREYC, but I know that is based on TV, and tries to overcome the staircaising artifacts by using some big modifications to the main algoirithm. The "problem", is that it has been designed for daylight, normal images, with gaussian noise. In that sense, our implementation of TGV is more flexible, since it works with linear or nonlinear data, and the use of the local support frame adds the capability to deal with poisson noise.
Also, from the user point of view, I think that TGV may lead to some "lost of detail" compared to GREYC, since the piecewise constant constrain enhances more sharp edges. But at the same time, TGV will look far more natural, with smooth gradients, and will avoid some of the typical "weird pixels" that arise near sharp edges with GREYC. IMHO, TGV is on top of the wave right now, and it should be even more powerfull in future releases. We have a major upgrade under development, and have several research lines already plotted.

Now, about when to use TGV best... The truth is that we don't know yet. I think that both steps combined will prove to be a good choise. I would use it first at linear stage to smooth a bit the data (low strengths), to facilitate the first stretch and intensity adjustments. Then, I would use as a finisher, to really smooth the background and low signal features (up to taste).

ACDNR based its philosophy on SGBNR And yes, we kept that in TGV too. Luminance/Chrominance separation is crucial to many denoising problems. Nevertheless, if in linear stage, I should probably try working on RGB first. Just thinking as a purist. But, of course, you all will help us find the best uses and applications of this new tool. 

[mdemita]

I cannot for the life of my find this tool in my release.  I am using 1.08.00.1015 (x64) RC7,  Any suggestions would be very welcome.

Thanks

Mike

[AstroScience]

Mike, did you saw that PI downloaded that update?
If so, go to Process/Modules/Install Modules/
on the window that will open click on "Search" , it should find TGVDenoise module.
Click on "Install" and it should work.
I don't know why it didn't that automatically as always but installing manually works.

[Geoff]

You could also try Resources>Updates>Check for Updates.
Worked for me.
Geoff