Jump to content

  •  

CNers have asked about a donation box for Cloudy Nights over the years, so here you go. Donation is not required by any means, so please enjoy your stay.

Photo

Wavelets in Registrax

  • Please log in to reply
10 replies to this topic

#1 DirtyRod

DirtyRod

    Sputnik

  • -----
  • topic starter
  • Posts: 41
  • Joined: 23 Mar 2021
  • Loc: Arizona

Posted 10 May 2021 - 11:46 PM

Anyone have any great articles or tutorials on adjusting Wavelets? Did my first couple of stacks and did decent for my first images but the wavelets seem to adjust different images differently so I have varying results depending on different planets I’m working on. Some of the articles I read suggest starting with layer 1 while others state that layer 1 is to capture noise so I should start with layer 6. I know it all depends on the image data but I’d like to understand what I’m doing versus just doing trial and error.

 

i did read the thread on magic vs science which went pretty dead so I’m looking mainly for a fairly simple how to for those of us that are just graduating from a point and shoot if such an animal exists.

Attached Thumbnails

  • 8738E88D-0715-4349-B124-20549474E06C.jpeg
  • 8F36772B-4001-425F-B823-081117980CF0.jpeg


#2 Tulloch

Tulloch

    Mercury-Atlas

  • *****
  • Posts: 2,812
  • Joined: 02 Mar 2019
  • Loc: Melbourne, Australia

Posted 11 May 2021 - 12:19 AM

Here's a tutorial - however, it's really just trial and error with the data you have, there's no magic formula :)

http://momilika.net/...6Processing.htm

 

Andrew


  • DSOs4Me and DirtyRod like this

#3 dswtan

dswtan

    Viking 1

  • *****
  • Posts: 617
  • Joined: 29 Oct 2006
  • Loc: Morgan Hill, CA

Posted 11 May 2021 - 01:49 AM

I agree with Tulloch. I seem to remember these tutorials (wwgeb - Planetary Imaging series) were helpful, if you like videos -- which works well in this case:

https://youtu.be/TsdoIRIFEe0

 

Something that helped me along was learning that the way to think about the wavelet layers is that they are like sharpening at different scales of granularity. Hitting the preview field (to the right of the slider) can help to show what's going on. It's different to something like regular photographic sharpening where it's more of a black box -- here, you get more control, at the expense of complexity.

 

Layer 1 is the finest granularity, and depending on your camera and settings, is likely down at the noise (grain) level, i.e. really fine structures in the image -- so there you are just sharpening noise, quite often. Then each layer is coarser and coarser in terms of what "scale" structures within the image are being sharpened.

 

You balance the sharpening you do at each level (i.e. scale of structure) with the denoise control per layer, because the sharper you make things, the more noisy it appears. It's a judgement call and you just have to tweak. It's very iterative.

 

But save yourself a few "schemes" as sets of these wavelet settings are call, and you can just call up a little library you'll create for your particular typical images. Personally I got to the point where I just found a recipe that worked for me for a particular object (like last year's Mars opposition for example), and I could just keep using the same scheme (per object). I have a few personal schemes for Jupiter, Saturn, solar, and the moon. Sometimes I might apply a scheme from a totally different object on a new object, at least as a starting point. My schemes are unique to my setup and imaging trains, so not so useful to others. But you can search the Internet for a few shared ones out there. 


  • DirtyRod likes this

#4 DirtyRod

DirtyRod

    Sputnik

  • -----
  • topic starter
  • Posts: 41
  • Joined: 23 Mar 2021
  • Loc: Arizona

Posted 11 May 2021 - 10:29 AM

I agree with Tulloch. I seem to remember these tutorials (wwgeb - Planetary Imaging series) were helpful, if you like videos -- which works well in this case:
https://youtu.be/TsdoIRIFEe0

Something that helped me along was learning that the way to think about the wavelet layers is that they are like sharpening at different scales of granularity. Hitting the preview field (to the right of the slider) can help to show what's going on. It's different to something like regular photographic sharpening where it's more of a black box -- here, you get more control, at the expense of complexity.

Layer 1 is the finest granularity, and depending on your camera and settings, is likely down at the noise (grain) level, i.e. really fine structures in the image -- so there you are just sharpening noise, quite often. Then each layer is coarser and coarser in terms of what "scale" structures within the image are being sharpened.

You balance the sharpening you do at each level (i.e. scale of structure) with the denoise control per layer, because the sharper you make things, the more noisy it appears. It's a judgement call and you just have to tweak. It's very iterative.

But save yourself a few "schemes" as sets of these wavelet settings are call, and you can just call up a little library you'll create for your particular typical images. Personally I got to the point where I just found a recipe that worked for me for a particular object (like last year's Mars opposition for example), and I could just keep using the same scheme (per object). I have a few personal schemes for Jupiter, Saturn, solar, and the moon. Sometimes I might apply a scheme from a totally different object on a new object, at least as a starting point. My schemes are unique to my setup and imaging trains, so not so useful to others. But you can search the Internet for a few shared ones out there.


Thanks. This one I watched and agree it was helpful. In fact, I think it helped me do my best image yet. I guess trial and error is it.4e50a29c04fc2c477734e363a504df09.jpg
  • dswtan likes this

#5 DirtyRod

DirtyRod

    Sputnik

  • -----
  • topic starter
  • Posts: 41
  • Joined: 23 Mar 2021
  • Loc: Arizona

Posted 11 May 2021 - 10:30 AM

Here's a tutorial - however, it's really just trial and error with the data you have, there's no magic formula :)
http://momilika.net/...6Processing.htm

Andrew


Thanks. I’ll Give it a watch.

#6 dcaponeii

dcaponeii

    Apollo

  • -----
  • Posts: 1,130
  • Joined: 01 Sep 2019
  • Loc: Waxahachie, TX

Posted 11 May 2021 - 04:03 PM

Based on all the tutoring I've received from folks on this forum in the past year I would strongly recommend several starting points.

 

1)  Linear - Don't use Dyadic or Linked wavelets.  The effects are much too strong.

2) Drizzle - Don't!  Unless you've got fantastic data drizzle is not your friend.

3) Layers - Use the Preview buttons to see which layer does what.  In general as you increase the layer number you are processing at a coarser scale.  Yes Layer 1 does most of the denoise you'll ever need it's also effective at sharpening.  I would suggest AT MOST you're using Layer 1 and Layer 2.  Every once in awhile I'll use Layer 3 but it's usually because the data is no good and I'm trying to squeeze something out of the noise.  ALWAYS fails!!

4) Strength of the layers not to exceed 0.140 (ish).  Denoise in Layer 1 as needed but not likely need to be above 0.35.  If you need that much denoise you're probably got too much strength in sharpening.

5) Play around with your own settings but as many have counseled me over these months complicated wavelet settings are usually a fools errand (my summary of lots of comments).  I'd suggest the K.I.S.S principle is the best path forward.

 

I'm sure I'm forgetting a lot of pearls I've received over the past year but I've hit my high spots.  The granularity in both your images is a result of too much sharpening and the denoise levels can't remove it all.  Back off on the sharpening across the board and see how your results turn out.

 

Hope this helps and definitely hope it doesn't hurt.


  • RedLionNJ, Kiwi Paul, DirtyRod and 1 other like this

#7 DirtyRod

DirtyRod

    Sputnik

  • -----
  • topic starter
  • Posts: 41
  • Joined: 23 Mar 2021
  • Loc: Arizona

Posted 11 May 2021 - 05:03 PM

Based on all the tutoring I've received from folks on this forum in the past year I would strongly recommend several starting points.

 

1)  Linear - Don't use Dyadic or Linked wavelets.  The effects are much too strong.

2) Drizzle - Don't!  Unless you've got fantastic data drizzle is not your friend.

3) Layers - Use the Preview buttons to see which layer does what.  In general as you increase the layer number you are processing at a coarser scale.  Yes Layer 1 does most of the denoise you'll ever need it's also effective at sharpening.  I would suggest AT MOST you're using Layer 1 and Layer 2.  Every once in awhile I'll use Layer 3 but it's usually because the data is no good and I'm trying to squeeze something out of the noise.  ALWAYS fails!!

4) Strength of the layers not to exceed 0.140 (ish).  Denoise in Layer 1 as needed but not likely need to be above 0.35.  If you need that much denoise you're probably got too much strength in sharpening.

5) Play around with your own settings but as many have counseled me over these months complicated wavelet settings are usually a fools errand (my summary of lots of comments).  I'd suggest the K.I.S.S principle is the best path forward.

 

I'm sure I'm forgetting a lot of pearls I've received over the past year but I've hit my high spots.  The granularity in both your images is a result of too much sharpening and the denoise levels can't remove it all.  Back off on the sharpening across the board and see how your results turn out.

 

Hope this helps and definitely hope it doesn't hurt.

Sounds like sage advice especially not overdoing it. My first few were great but then I kept pushing the levels and the image deteriorated in front of my eyes. I’ll try taking off linked as well. One person advised leaving it on so I did that but I do understand it multiplies the effects as up move through the layers so perhaps it aligns to not overdoing it.

 

appreciate the tips.



#8 Kiwi Paul

Kiwi Paul

    Viking 1

  • -----
  • Posts: 758
  • Joined: 13 Jul 2020
  • Loc: Carterton, New Zealand

Posted 11 May 2021 - 06:09 PM

Dcaponeii you have hit the nail on the head with your observations above. Your summary is exactly what I find.
I find I start out with the wavelets all set to their initial positions, load the image and then move the 1st slider all the way over to the right and then increase the value judiciously and stop when the image starts to look too sharpened - often above 0.15 to 0.20, maybe back off a bit. Maybe use a little slider #2. Then advance the denoise on slider #1 until speckling is muted out, then I back off the denoise until the noise just starts to reappear. As Andrew says, how much noise you tolerate is a personal choice. I can tolerate just a little because I think it preserves the finest detail you may have. I save these (and gamma) settings and use them for subsequent images for the same imaging session (with tweaks).
I find I usually can’t use these settings for other imaging sessions because the capturing conditions are usually different
Cheers Paul
  • RedLionNJ, dcaponeii and DirtyRod like this

#9 RedLionNJ

RedLionNJ

    Skylab

  • *****
  • Posts: 4,314
  • Joined: 29 Dec 2009
  • Loc: Red Lion, NJ, USA

Posted 11 May 2021 - 06:26 PM

This (additional) advice is not directly wavelet-related, so feel free to discard if it's not useful:

 

Don't expect every capture to be worthy of even processing.  This is a HUGE difference from other facets of astrophotography.  We're trying to work at image scales measured in fractions of an arcsecond here.  Frequently the sky itself doesn't cooperate ("clear" is just not enough) and limits the recordable resolution to scales measured above an arcsecond. When conditions are like that, no amount of sharpening is going to bring out genuine detail - it just wasn't captured in the first place. As you start out, you may find as many as four out of five (or nine out of ten!) nights are like this - planetary work is futile under those conditions.

 

Your stack also needs to have a really good signal-to-noise ratio in order for wavelets to be really effective - else you end up amplifying noise as well as signal and that never ends positively.

 

The ideal pre-wavelet stack looks like a thin mist is covering the fine details. Applying just the right wavelets to that has the effect of gathering the mist up onto the higher-contrast details where the photons should have been recorded in the first place, had capture conditions been perfect.

 

Good luck out there!


  • dswtan, Tulloch, dcaponeii and 3 others like this

#10 DirtyRod

DirtyRod

    Sputnik

  • -----
  • topic starter
  • Posts: 41
  • Joined: 23 Mar 2021
  • Loc: Arizona

Posted 11 May 2021 - 09:07 PM

This (additional) advice is not directly wavelet-related, so feel free to discard if it's not useful:

 

Don't expect every capture to be worthy of even processing.  This is a HUGE difference from other facets of astrophotography.  We're trying to work at image scales measured in fractions of an arcsecond here.  Frequently the sky itself doesn't cooperate ("clear" is just not enough) and limits the recordable resolution to scales measured above an arcsecond. When conditions are like that, no amount of sharpening is going to bring out genuine detail - it just wasn't captured in the first place. As you start out, you may find as many as four out of five (or nine out of ten!) nights are like this - planetary work is futile under those conditions.

 

Your stack also needs to have a really good signal-to-noise ratio in order for wavelets to be really effective - else you end up amplifying noise as well as signal and that never ends positively.

 

The ideal pre-wavelet stack looks like a thin mist is covering the fine details. Applying just the right wavelets to that has the effect of gathering the mist up onto the higher-contrast details where the photons should have been recorded in the first place, had capture conditions been perfect.

 

Good luck out there!

Good stuff. I definitely understand that some of my captures are not worth the effort after wasting time on them. I’ll take the noise ratio into consideration going forward.

 

Rod.



#11 Sarciness

Sarciness

    Sputnik

  • -----
  • Posts: 31
  • Joined: 15 Apr 2021
  • Loc: Suzhou, China

Posted 20 May 2021 - 09:08 PM

Based on all the tutoring I've received from folks on this forum in the past year I would strongly recommend several starting points.

 

1)  Linear - Don't use Dyadic or Linked wavelets.  The effects are much too strong.

2) Drizzle - Don't!  Unless you've got fantastic data drizzle is not your friend.

3) Layers - Use the Preview buttons to see which layer does what.  In general as you increase the layer number you are processing at a coarser scale.  Yes Layer 1 does most of the denoise you'll ever need it's also effective at sharpening.  I would suggest AT MOST you're using Layer 1 and Layer 2.  Every once in awhile I'll use Layer 3 but it's usually because the data is no good and I'm trying to squeeze something out of the noise.  ALWAYS fails!!

4) Strength of the layers not to exceed 0.140 (ish).  Denoise in Layer 1 as needed but not likely need to be above 0.35.  If you need that much denoise you're probably got too much strength in sharpening.

5) Play around with your own settings but as many have counseled me over these months complicated wavelet settings are usually a fools errand (my summary of lots of comments).  I'd suggest the K.I.S.S principle is the best path forward.

 

Thanks! I might try with these recommendations next time I'm imaging and see what I can do.
 




CNers have asked about a donation box for Cloudy Nights over the years, so here you go. Donation is not required by any means, so please enjoy your stay.


Recent Topics






Cloudy Nights LLC
Cloudy Nights Sponsor: Astronomics