Author Topic: multiscale processing ?  (Read 5130 times)

Offline wvandenberghe

  • Newcomer
  • Posts: 27
multiscale processing ?
« on: 2011 April 28 10:28:55 »
Hello Vicent


I have read about multiscale integrating  to extract structures for dust motes that cannot be elliminated with a normal flat frame.

Discussion of http://pixinsight.com/examples/NGC6914-CAHA/en.html

I have no idea how to implement this  practically  this with pixinsight.
atrouswavelettransform ?  how do you integrate both sets of images ? pixelmath ?
Can you get me into the right direction?

Clear Skies

Wolfgang Van den Berghe

homepage : www.vstobservatory.net

Offline vicent_peris

  • PTeam Member
  • PixInsight Padawan
  • ****
  • Posts: 988
    • http://www.astrofoto.es/
Re: multiscale processing ?
« Reply #1 on: 2011 April 29 01:19:22 »
Hi Wolfgang,

For this process, I usually make two different image integrations. The first will be the usual integration (don't care if you don't reject perfectly the dust donuts or the artificial satellites).

For the second image integration process, I make a set of images without the first 2 - 3 wavelet layers. This is done with an image container by applying an ATWT instance to the image set. Then I make the integration of this new set of images; here you can be more restrictive in order to reject all the artifacts.

Last we must all the data together. We will put the first 2 - 3 wavelet layers of the resulting image from first integration process over the resulting image of the second integration. We will do this in PixelMath:

Int1 - Int1_LS + Int2

Where:

Int1 is the resulting image from the classical image integration process.
Int1_LS is Int1 but without the first 2 - 3 wavelet layers.
Int2 is the resulting image from the second image integration process run.

Here you can have a problem: Ringing. This comes from rejection differences in some structures in the both image integrations (for example around bloomings, where you have very sharp edges). This can be solved if we use another PixelMath equation:

iif ( Int1-- (Int1 - Int1_LS + Int2) > k, Int1, Int1 - Int1_LS + Int2 )

In this conditional equation, we will put only Int1 in the final image if the difference between the classical integration and the multiscale integration is higher than k; if not, we will put the multiscale integration result. As we are working with linear images, usually values of k are very low (tipically in the order or >0.01). You can play with this value if you get some ringings from the first equation.

As an alternative, we can use a variable. In the formula field, put first the variable declaration and then the conditional formula:

x = Int1 - Int1_LS + Int2; iif ( Int1 -- x > k, Int1, x )

In the Symbols field you must put x to tell PixelMath that x is a variable. This formula is much faster to calculate than the prior one.

We are going to implement in the very near future this kind of integration as a script.


Good luck!
Best regards,
V.

PD: Remember to always uncheck the Rescale checkbox.
« Last Edit: 2011 April 29 01:59:21 by vicent_peris »

Offline mmirot

  • PixInsight Padawan
  • ****
  • Posts: 881
Re: multiscale processing ?
« Reply #2 on: 2011 April 29 06:25:25 »
Requires a large dither ?

Max

Offline wvandenberghe

  • Newcomer
  • Posts: 27
Re: multiscale processing ?
« Reply #3 on: 2011 April 30 02:05:39 »
Hi Vicent

Thank you very much for your detailed reply.
I tried it out  on my image of the Eyes galaxies , but don't see much improvement.
Maybe its because I have not dithered my exposures, because I find star shapes are elongated if I try that.
I may need to do a lot of trial and error with Atrouswavelet to get rid of the big dust particle on the Baader  luminance filter.
Better to blow it off in the first place.
See screenshot attached made by RCOS  RC 14,5inch f/9 Camera SBIG STL11000M for the original unmodified image, only DBE applied.

Best regards

Wolfgang Van den Berghe
www.vstobservatory.net



Offline Nocturnal

  • PixInsight Jedi Council Member
  • *******
  • Posts: 2727
    • http://www.carpephoton.com
Re: multiscale processing ?
« Reply #4 on: 2011 April 30 11:19:21 »
Hi,

I'm curious why dithering would elongate your stars. I dither all the time and I've never seen that. Provided each sub has round stars and registration works OK I see no reason the stacked image would have elongated stars. Could you show an example?
Best,

    Sander
---
Edge HD 1100
QHY-8 for imaging, IMG0H mono for guiding, video cameras for occulations
ASI224, QHY5L-IIc
HyperStar3
WO-M110ED+FR-III/TRF-2008
Takahashi EM-400
PIxInsight, DeepSkyStacker, PHD, Nebulosity

Offline wvandenberghe

  • Newcomer
  • Posts: 27
Re: multiscale processing ?
« Reply #5 on: 2011 May 01 00:44:50 »
Sander
I have used Autodither  for CCDSOFT , the mount is a Paramount ME in DirectGuide autoguiding mode.
The statistics clearly show that star shapes are degraded when dithering is used.

Example without dithering :
http://www.vstobservatory.net/gallery/galaxy/00000027.NGC_2683.jpg

with dithering :
http://www.vstobservatory.net/gallery/galaxy/00000002.NGC_2683.jpg

CS

Wolfgang

Offline Nocturnal

  • PixInsight Jedi Council Member
  • *******
  • Posts: 2727
    • http://www.carpephoton.com
Re: multiscale processing ?
« Reply #6 on: 2011 May 01 01:20:57 »

Wolfgang,

I guess the conclusion is that CCDSoft's dithering is broken. Using Nebulosity and  PHD this doesn't happen. Dithering should be done between exposures and thus should have no effect on quality of the lights.
Best,

    Sander
---
Edge HD 1100
QHY-8 for imaging, IMG0H mono for guiding, video cameras for occulations
ASI224, QHY5L-IIc
HyperStar3
WO-M110ED+FR-III/TRF-2008
Takahashi EM-400
PIxInsight, DeepSkyStacker, PHD, Nebulosity

Offline wvandenberghe

  • Newcomer
  • Posts: 27
Re: multiscale processing ?
« Reply #7 on: 2011 May 01 01:47:00 »
I don't see any reference in Nebulosity or phd to Paramount ME direct guiding support..

I guess my options are CCDSOFT no dithering or
MaximDL, I haven't tried yet with my setup.
 CS
Wolfgang

www.vstobservatory.net

Offline Nocturnal

  • PixInsight Jedi Council Member
  • *******
  • Posts: 2727
    • http://www.carpephoton.com
Re: multiscale processing ?
« Reply #8 on: 2011 May 01 11:27:09 »

I don't know how P-ME guiding works but if it has a guide port (ST-4) and if your guide camera is supported by PHD then PHD can guide it.

In any case, this is a problem specific to your hardware/software combo. What do fellow P-ME/CCDSoft users say about it?
Best,

    Sander
---
Edge HD 1100
QHY-8 for imaging, IMG0H mono for guiding, video cameras for occulations
ASI224, QHY5L-IIc
HyperStar3
WO-M110ED+FR-III/TRF-2008
Takahashi EM-400
PIxInsight, DeepSkyStacker, PHD, Nebulosity