Author Topic: how to separate large scales and small scales and rebuild the image  (Read 6350 times)

Offline lucchett

  • PixInsight Old Hand
  • ****
  • Posts: 449
Hi,
I'd need help to see if I understand well the process.

separation:
one option would be
- to use the MLT, using linear and canceling higher levels (let's say above 4)
-to use the MLT, using linear and cancelling lower levels (1-3)?

rebuild:

??

Thanks a lot!
Andrea

Offline tgeib

  • Newcomer
  • Posts: 5
Warren Keller's tutorials have you remove the residual layer from a clone and then subtract from the original image to produce a small and large scale image.  Experiment with the number of layers to get the correct division between small and large scale structures.

Clear skies,

Tim Geib

Offline tgeib

  • Newcomer
  • Posts: 5
One more thing.  He used Dyadic with ATrousWavelet Transform.

Tim

Offline lucchett

  • PixInsight Old Hand
  • ****
  • Posts: 449
Thank you Tim.
this worked well.

I wonder what could be the process to extract a single layer and recombine it later.

I'd like to understand why your Lum-LS=SS is different from what I get isolating the first n layers with a multiscale tool.

I am definetely missing how wavelets works...


Andrea

Offline tgeib

  • Newcomer
  • Posts: 5
Andrea,
What I do is actually a little different.  If Lum is your original image, Lum - SS = LS, where SS is the clone of Lum with the Residual layer removed.  There's not much difference from what you're doing, but it can be a bit faster, depending on long it takes your computer to cancel each wavelet.

The subtraction occurs with pixel math.  The rebuild does as well. When you're done working on each scale, add them back together. Consider using the rescale result option when reassembling, since you've probably created different changes in each scale.

If you're not familiar with using pixel math, I learned a lot from Gerald Wechselberger's page:  http://www.werbeagentur.org/oldwexi/PixInsight/PixInsight.html

Tim


Offline CharlesW

  • PixInsight Enthusiast
  • **
  • Posts: 87
Another possibility and one I've just found is Extract Wavelet Layers in Script/Image Analysis. You tell it how many layers you want and let 'er rip. Of course it would be nice if there was a recombine script but not yet.

Offline jkmorse

  • PixInsight Padawan
  • ****
  • Posts: 931
  • Two questions, Mitch . .
    • Jim Morse Astronomy
Great stuff!, just tagging for future reference.

Thanks,

Jim
Really, are clear skies, low wind and no moon that much to ask for? 

New Mexico Skies Observatory
Apogee Aspen 16803
Planewave CDK17 - Paramount MEII
Planewave IFR90 - Astrodon LRGB & NB filters
SkyX - MaximDL - ACP

http://www.jimmorse-astronomy.com
http://www.astrobin.com/users/JimMorse

Offline Harry page

  • PTeam Member
  • PixInsight Jedi Knight
  • *****
  • Posts: 1458
    • http://www.harrysastroshed.com
Hi

For the lazy like me this script works very well

http://pixinsight.com/forum/index.php?topic=8079.0


Regards

Harry
Harry Page

Offline Warhen

  • PTeam Member
  • PixInsight Old Hand
  • ****
  • Posts: 490
    • Billions and Billions
In regards to Tim's comment, while creating the small-scales image first is easier (turning residual off), it may not always result in the smoothest large-scales image. In some scenarios, creating the large-scales image first (turning off all small scales with residual on) yields the better result!
Best always, Warren

Warren A. Keller
www.ip4ap.com

Offline lucchett

  • PixInsight Old Hand
  • ****
  • Posts: 449
Good info, thank you all.
I understood how to separate large scale and small scale.
what I am still missing is how to combine single layer + residual to rebuild an image (not really needed but useful to understand how it works).
also different tools seem to give diffrent results (MLT, ATWT, script), but I think this are variation of teh same concept.

Andrea

Offline msmythers

  • PTeam Member
  • PixInsight Jedi
  • *****
  • Posts: 1178
    • astrobin
Andrea

The simplest is just PixelMath. Here I separated the first 8 layers for the Small image. Then I separated the residual layer for the Large image. I used a simple Small+Larger expression in PixelMath to create the New combined image. There are probably better ways but you get the idea.


Mike

Offline lucchett

  • PixInsight Old Hand
  • ****
  • Posts: 449
Hi Mike,
this works if you split the image in two:

-IMG1 is the original
-SS: disabling the residual
-LS: IMG1-SS

IMG1: LS+SS

My question is:

using a tool like MMT, I can have a SS image disabling the residual layer AND a Residual layer.
is it possible to combine them in some way, to rebuild the original image?

if this is what you did, please tell me how do you separate the two layers.

Thank you!
Andrea

Offline msmythers

  • PTeam Member
  • PixInsight Jedi
  • *****
  • Posts: 1178
    • astrobin
Andrea

This is all I did to generate the Small and the Large.


Mike

Offline lucchett

  • PixInsight Old Hand
  • ****
  • Posts: 449
Ok, I did exactly what you did.

using "linear" does help a lot but if you compare the resulting image to the original one, they are different.
the new one lacks details

can you confirm?



Offline msmythers

  • PTeam Member
  • PixInsight Jedi
  • *****
  • Posts: 1178
    • astrobin
Try it without rescale result and create a new image instead of replace target. This is exactly what I did.

Mike