›

masked stretch issue after TGV denoise

Benoit Saintot
in PixInsight
Hello to all.

I'm having difficult time to use masked stretch after TGV Denoise + MMT

When I do a masked stretch on my Lum (after DBE and Decon ) it looks ok.
But if I run a masked stretch after a TGV / MMT denoise, it gives me weird background. 
I tried to fine tune the mask,  adding local support,  fiddeling with smoothness and edge protection, running MMT then after to get rid of medium  / low freq noise remaining, but it always lead to some weird background, like IFN ( and for sure that's not !)

So, as i was wondering if it was not only a display issue, and backed up a little bit, and applied the same stretch to bothe my denoise lum and my undenoised one. The difference is more subtle ( and actually pops the medium / low noise out with the masked stretch )

Attached are three screen captures to illustrate this .
As it might be difficult to see after jpg compression, and as it is way more visible after an EZ denoise process, I also attached a masked stretch after an EZ denoise run.

So, I'm really wondering what's happening there. Whats is intrisically leading to this background inside masked stretch after the denoise process ?
And what should I do to be able to run a masked stretch after a linear denoise ? 

Thanks a lot for your help !

Comments

  • BlockHeadBlockHead
    Hi Benoit,

    This is a perennial issue. (People complain about this more commonly with DBE.)
    The weird background you see in the masked stretch result *is really there all along*. You are just not seeing it easily in the other image. Masked stretched (and a number of other tools) raises the background brightness (you can control this) enough that when the new STF is applied the histogram will show everything in the image. If you brighten your original image with enough contrast... you can prove to your self the issue was there all along.

    Applying the same stretch STF (auto) stretch does not work. The images have completely different histograms and will be displayed with different degrees and brightness and contrast you apply the same settings to both. This is another common misconception. You need to remember when you applied masked stretch (or other things) you really did change the numbers!

    There isn't a simple way to match the settings...but you can adjust each so that you are seeing the same brightness/contrast level...and you will find the same stuff is in the original.

    Now your next question is... how do I avoid this noisy background. The answer comes in two points.

    1. Your data does not allow you to display the faintest signal (without seeing the noise). There is always a limit for all images/data. This is the threshold of useful information. You want to include the faintest signal? Buy a better camera, a bigger telescope, a darker sky, take longer exposures... This applies at all levels of imagery...there is a limit to what you can show. 

    2. The noise in your image is somewhat correlated. I mean to say... it isn't just pixel level noise. This is what makes it hard to correct. There are correlations larger than just 1-2 pixels. Better flat fielding might help...but it is very very hard to do. My data from the 0.8m Schulman telescope was not any better than what you are showing. There are some fancy involved ways of massaging the background with masks and noise generation..etc etc. But I don't think this is ultimately how you want to spend time processing images... where it takes effort for every image you take to wrangle this noisy stuff.

    -the Blockhead
  • Benoit Saintot
    Hello Adam

    Thanks a lot for your reply.
    Si if I do understand, the masked stretch does only reveal what is already in the picture, these artifacts were only hidden by the less "aggressive" standard HT.
    And the further we push the noise reduction, the more artifacts like this MS could reveal...

    One thing that's not still clear to me, is what in the denoise process makes that the stretch after ( HT, STF, MS,...) will reveal this.
    In other terms, why, after a denoise, a stretch reveal more artifacts?

    Finally, is there a way, with these data, to tame a bit Masked Stretch as to have it do it's job on the bright parts ( M51), but not that much on the background? I fear a " mask masked stretch" would lead to non uniformity.

    Thanks a lot

    Ben
Sign In or Register to comment.