Release 1079 STF function behavior

[niteman1946]

Hi Group,

I posted earlier in the Release Information section regarding Masked Stretch behavior both before and after application of TGVdenoise and Deconvolution (everything in linear form).
http://pixinsight.com/forum/index.php?topic=6641.0
I asked why Masked Stretch would produce so much more intense stretch after these two operations than before them.

Now I see that Screen Transfer Function (STF) exhibits the same characteristics for before and after TGVdenoise alone.  The attached picture shows three shots of the same file (all linear).
Upper Left -  after DBE, autostretched with STF
Upper Right -  immediately after TGVdenoise with NO further autostretch.
Lower Center -  same as upper right, but after STF is "cleared" and then re-imposed

Prior to the recent round of PI new revision releases I can say with confidence that TGVdenoise did not cause the autostretch to become more intense.
I can still adjust the sliders in STF to present a more pleasing image while I work, but just curious as to why this change.

Thanks,
Mark

[sreilly]

My question to this is should TGVD be used on linear data? I haven't used it at that point and usually use it at the last part of my processing and the data is non-linear then. I'd try that and see if you get the same results.

[Carlos Milovic]

I think that the answer is that TGVDN is changing the intensity distribution of the pixels, so the auto STF calculates another set of parameters that is more aggressive than in the previous step. So, contrast in the background is exaggerated, while in the galaxy it is lost.

[niteman1946]

Hi Steve,

It appears that TGVDenoise continues to work well in linear mode.  I prefer to use it there.  It's just that the more intense autostretch after Denoise is a new behavior in  release 1079 (do not remember noticing it in release 1071).

Mark

[niteman1946]

I think that the answer is that TGVDN is changing the intensity distribution of the pixels, so the auto STF calculates another set of parameters that is more aggressive than in the previous step. So, contrast in the background is exaggerated, while in the galaxy it is lost.

Thanks Carlos,
I suppose it's really a trivial point.  I just remember this behavior prior to release 1079.
Mark

[niteman1946]

I think that the answer is that TGVDN is changing the intensity distribution of the pixels, so the auto STF calculates another set of parameters that is more aggressive than in the previous step. So, contrast in the background is exaggerated, while in the galaxy it is lost.

Thanks Carlos,
I suppose it's really a trivial point.  I just remember this behavior prior to release 1079.
Mark

My mistake,
I meant to say "I just don't remember this behavior prior to release 1079."
Has anyone else noticed this?

Mark

[Carlos Milovic]

The way the statistics are calculated changed in the last releases, so this may explain the differences.

[niteman1946]

Thanks Carlos,

Mark