ATrousWaveletTransform question (how to apply and keep it)

[steven_usa]

To help clarify something up for me.... Since I suspect I'm not understanding something about STF:

Suppose I have a stacked/calibrated image, which I then apply BDE and SCNR while the mage is STF Auto Stretched.  Typical standard stuff...

Now to clear up some noise, suppose I apply ATrousWaveletTransform while the image is stretched.  I set some "Noise Reduction" parameters that seem to work, and press Apply.

Now suppose the image looks GREAT, that's what I want to keep, don't do anything else.  How do I keep what I am seeing? (well, other than doing a screenshot and pasting a dumb BMP somewhere)

What I mean is, if I save the image as a FITS or remove the STF and re-apply the auto stretch, the image looks like trash again (well, not that bad, but not nearly as good as it did with the original stretch and the ATrous applied).   What I usually have to do is remove any auto STF, then apply something like MaskedStretch and a few other steps and sometimes get close to what I wanted with the original auto stretch and ATWT.   

Is that just how it is? or am I missing something?


Here are 3 screenshots to try to show what I mean:
#1) original Auto STF
#2) ATWT applied with 4 noise reduction (looks smooth and nice, I like it)
#3) Turn off Auto STF and redo the stretch.  It's nice and all, but fuzzy again...

[cdesselles]

Hmm I'm no expert but, why don't you try this: 

Do your auto stretch and AWTF getting the pic the way you like it again. Then, open the Histogram Transformation Tool and drag the STF "new instance" icon to the bottom bar of the HTT.  Turn off STF then apply the HTT to your image.  This should transfer all of the STF (linear state) settings to your image in a non-linear permanent state.  Then you can save it as a FITS, TIFF, JPEG or whatever you wish.

(PS: Let me know how well this works for you.)
 

[pfile]

yes chris is right. the STF settings can be copied over to HT at any time, and then you can apply the HT to the image to make that STF "permanent".

the phenomenon that you're seeing is that STF is looking at the statistical properties of the image in an attempt to calculate a stretch that "makes sense". after the noise reduction, it sees different statistics and so computes a different stretch - by the looks of it a lot more aggressive.

generally speaking i think most people do the HT stretch "by hand" just to have control over the process. you can turn on the real-time preview when you are messing with the HT sliders to see what the effect will be on your image. don't forget to turn off the STF while you're doing that though, or you'll see a "double stretch" in the real-time preview.

rob

[steven_usa]

Thanks, that was exactly it   Yes, it did work, confirmed.

So honestly, I never used the HT much.  My histograms are so bad (I think), they are very near the left side, and I didn't even notice the sliders.  So the precision of the sliders seems awkward to me, since I have zoom so far in to even reveal the histogram. 

Thanks again!   So, maybe I don't need to do MaskedStretch anymore...

1 more question: when do I NOT want to use the 24-bit lookup table during the auto STF?  I just keep it on all the time.

[pfile]

MaskedStretch is actually a superior tool, but usually you still need to do more "regular" HT or curves when it's done. so you can keep using MT if you like

24-bit STF was added because in some cases the regular STF would show posterization artifacts. this usually happens when you have a 64-bit image, such as you get out of the HDRComposition process. for 16-bit and 32-bit images normal STF is usually fine. 24-bit STFs eat a lot of memory so it's probably best not to use them unless it's really necessary. but it won't harm anything to use 24 bit STF.

sounds like you need to increase your subexposure lengths?

rob

[steven_usa]

Ah thanks, ok when using super bias I just got the habit of using 24-bit STF to view them (and it made a difference there).  You're probably right that it is otherwise not necessary (for me).

I was doing CurveTransformation after the MaskedStretch.   And I'll still experiment with the MaskedStretch parameters.  It's just, I don't think it has a Preview?  But I guess we can always apply and undo it.

Weird thing is my exposures were still 5-10 minutes.  And actually, it's just on the galaxy and dark things (like this Iris) where the histograms are poor.  It probably wasn't wise to try to process those more difficult images first.

PI is pretty awesome.  I think my only nit so far is that during BPP, when selecting a reference frame, I can't do other functions while using BPP.  So if I forget to choose a reference first, I have to cancel out of the entire dialog and start over.  Or, I might use some other tool like DSS to try to pick which frame I want to use as the reference.

Another nit might be that, I've gotten updates two times so far, and it seems after both updates all my tool settings got reset back to their default values.   That's ok, it's just something to be aware of and remember.  You kind of take the settings for granted, and might not even do a stretch to observe things as you're going, and took me awhile to realize what had happened when I suddenly wasn't getting same results anymore.
EDIT: Specifically, the ImageIntegration "Rejection algorithm" seems to get reset from "Winsorized Sigma Clipping" back to "No rejection" during an update.

[Don]

Hi Steve,

You can save the current state of the BPP screen if you need to get out and come back in.  Just drag the triangle at the bottom left of the BPP panel onto an empty area of the desktop to create a process icon, then exit from BPP.  When you double-click the process icon, it brings up a script window.  Click the "Apply Global" button (the circle at the bottom of the panel) and it will re-create your BPP panel just as it was when you created the process icon.

Your images do indeed have a very dark background -  I looked at your 10 minute subs of M101 (800 ISO), and there is no space between the left end of the histogram and the sky background hump.  You must have shot them in a VERY dark location.

Don

[steven_usa]

Yep, Three Rivers Foundation, Comanche Springs Observatory, perhaps the darkest site there is in Texas (with power/shower facilities), about 2-3 hours west of Wichita Falls.    The milky way is visible from horizon to horizon. So dark (on non-moon nights), you can see your shadow being cast from the milky way itself.  Borderline between Bortle 1 and 2, whereas my backyard is a solid red zone.   Fort Davis I hear has less facilities, but higher altitude; though it's 8 hours from me and I haven't ventured there yet.   Fort Griffen is also very nice, next to a small state park also with power/shower facilities.

Thanks for the BPP tip!  I didn't fully grasp the utility of those instance tabs yet, but I'm getting there

[pfile]

well not being able to get past being read-noise limited in your backgrounds is definitely a good problem to have!

i think juan has wanted to fix the javascript stuff to be non-modal for some time now, but i suppose it's probably a pretty hard problem. he'll get to it 

rob