Issue with TGVDenoise and Stretch

[SteveJP]

Hi, There is something I do not understand or are doing wrong. When I process a linear image with TGVDenoise, I apply various amounts of Strength, Edge protection etc, and when I think I have a satisfactory result, I find that re-applying the STF AutoStretch results in a different image compared to just toggling the stretch on and off with F12.

To explain in more detail.

1 - I start off with linear image (pretty dark of course, so you can't see much)
2 - STF Autostretch applied so I can see the affect of TGVDenoise with various parameters
3 - TGCDenoise is applied to the image - Noise much reduced - Great
After TGV Denoise I can toggle stretch on an off with F12 and image toggles between dark unstretched image and stretched image with noise reduced
4 - BUT, if I re-apply STF Autostretch, I get a results that is way different to what I saw when I applied TGVDenoise to the image whilst I had the autostretch turned on.

Seems I can't just try different TGVDenoise parameters without re-applying stretch every time. Should that be the case? What am I not understanding here. Same thing happened when I was applying the TGV to pre-view images. The previewed result, when was not the same as the result after re-applying the default stretch.

Any guidance would be most welcome.

[bulrichl]

Hi Steve,

F12 will disable the ScreenTransferFunction (STF). Please use Ctrl+Z for undoing, and Ctrl+Y for redoing the last process.

Bernd

[SteveJP]

Hi Steve,

F12 will disable the ScreenTransferFunction (STF). Please use Ctrl+Z for undoing, and Ctrl+Y for redoing the last process.

Bernd

Hi Bernd,
I guess the main problem is that if you are working with a previews of the image with the stretch applied, you can't rely on what you see unless you undo and re-apply the stretch. I haven't found this with other uses of preview, generally previews of the stretched image are accurate.
Thanks
Steve
 

[RickS]

STF uses the image statistics to calculate the autostretch.  After noise reduction the changed stats will give you a more aggressive autostretch.  I think that's probably what you're seeing.

[SteveJP]

STF uses the image statistics to calculate the autostretch.  After noise reduction the changed stats will give you a more aggressive autostretch.  I think that's probably what you're seeing.

Yes, that sounds right. Inconvenient though, kind of defeats the purpose of the preview. I haven't seen this mentioned in demos of TGV I've watched where the params have been tweaked whilst using the pre-view.
Thanks Rick

[AstroDude]

Very same thing happens  with MSHTrans.. or MMT  for denoising..you apply it them you re- stretch  ( STF)  then  BOOOM  worse noise then ever..

[pfile]

it is not worse noise than ever - it is just that in the face of smoother image statistics, STF auto-computes a stretch that is too aggressive. this is what rick just said above.

[Geoff]

I often save an instance of the screen stretch that was applied before the noise reduction. I can the use it at any time afterwards if I don’t want the more aggressive stretch.

[SteveJP]

I often save an instance of the screen stretch that was applied before the noise reduction. I can the use it at any time afterwards if I don’t want the more aggressive stretch.

Geoff/pfile, I have no doubt this explanation is correct. What initially confused me what that turning the stretch on and off with F12 was not the same thing as pressing the stretch button when looking at the unstretched image, however in most other cases it has the same result, albeit with possible other cases as AstroDude pointed out. I see now that there is a difference between reapplying the original stretch and re-computing it. Not obvious to a new user, or even perhaps a not so new user.
Thanks.