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

Drizzle and Deconvolution finally paying off - Reworks: Mars, Jupiter & Saturn

  • Please log in to reply
15 replies to this topic

#1 leoyasu

leoyasu

    Explorer 1

  • -----
  • topic starter
  • Posts: 64
  • Joined: 03 May 2013
  • Loc: Buenos Aires, Argentina

Posted 23 October 2020 - 08:54 PM

Hi everyone! This time I don't have new photos but I have reprocessed old ones. After some tips I started to drizzle my videos and then try to use deconvolution instead of wavelets or at least a mix. The old images in question come some from these posts: Jupiter + NTrZ Storms 11/09 - Jup + Storms, Sat (IR) and Mars 14/09 - SW14" dob + QHY462. Back then I only processed my images in native resolution and then sharpen only with wavelets and in PS with unsharp mask. Now I apply x1.5 drizzle (with x3 artifacts start to emerge for now) and then I apply deconvolution. Finally and after derotating I do a few iterations with Smart Sharpen on PS. The differences are more noticeable on Mars but in general I really liked the results.

 

1 for non-drizzle data and RS6 wavelets sharpening

 

2 for x1.5 drizzle data and deconvolution sharpening

 

 

2020-09-15-0106_2-LY-Jup.jpg

 

2020-09-15-0106_2-LY-Jup_RGB_Drizzle15_80%.jpg

 

 

2020-09-15-0232_9-LY-Sat2.jpg

 

2020-09-15-0232_9-LY-Sat_RGB_Drizzle15.jpg

 

 

2020-09-15-0545_1-LY-Mars4.jpg

 

2020-09-15-0545_1-LY-Mars_RGB_80%_WJ2.jpg

 

 

2020-09-29-0436_8-LY-Mars_RGB.jpg

 

2020-09-29-0436_8-LY-Mars_RGB_80%.jpg

 

 

I have a bad habit of deleting everything that I don't use, so I'm looking at which sessions I had the original videos left to stack again but at the moment these were the results of the experiments. I think I'm on a better path!

 

Best regards, Leo.

 

 

 

 


  • Sunspot, JMP, Achernar and 34 others like this

#2 Tyson M

Tyson M

    Fly Me to the Moon

  • *****
  • Posts: 5,044
  • Joined: 22 Jan 2015
  • Loc: Canada

Posted 23 October 2020 - 08:59 PM

Outstanding images! Excellent processing.



#3 Foc

Foc

    Viking 1

  • -----
  • Posts: 999
  • Joined: 17 Nov 2016
  • Loc: South Canberra

Posted 23 October 2020 - 11:40 PM

Looks like it has really paid off with your Mars images!



#4 ryanha

ryanha

    Messenger

  • -----
  • Posts: 432
  • Joined: 05 Aug 2020

Posted 23 October 2020 - 11:44 PM

gorgeous images!



#5 DarkMain

DarkMain

    Lift Off

  • -----
  • Posts: 18
  • Joined: 04 Aug 2020

Posted 24 October 2020 - 04:19 AM

What program are you using for deconvolution?



#6 Jkaiser3000

Jkaiser3000

    Vostok 1

  • *****
  • Posts: 126
  • Joined: 22 Apr 2020
  • Loc: Caracas

Posted 24 October 2020 - 07:37 AM

Wow, it’s all I can say. There’s a striking difference between both approaches and, option 2 is always better, sometimes by a mile 

 

I would also like more information on how you achieved these, if you don’t mind


  • leoyasu likes this

#7 Yu Gu

Yu Gu

    Vostok 1

  • *****
  • Posts: 179
  • Joined: 18 Jun 2006

Posted 24 October 2020 - 07:45 AM

Wow! I have the same questions as the posts above.



#8 leoyasu

leoyasu

    Explorer 1

  • -----
  • topic starter
  • Posts: 64
  • Joined: 03 May 2013
  • Loc: Buenos Aires, Argentina

Posted 24 October 2020 - 11:45 AM

Thank you all for the posts and likes! waytogo.gif

 

For deconvolution I was experimenting with AstraImage, Fitswork and AstroSurface but I got more consistent results with AstraImage. The resulting image contained less noise, graininess and a bit more sharpness. I use the Lucy-Richardson deconvolution on the luminance channel. In particular, I apply between 10 and 30 iterations based on the data quality with a strength between 5 and 10. Then for the Blur kernel I use a size between 1.5 and 2. I'm pretty new to this, so I'm still gaining experience.

 

Leo.


Edited by leoyasu, 24 October 2020 - 11:46 AM.

  • Yu Gu, Tulloch and DarkMain like this

#9 Tulloch

Tulloch

    Vanguard

  • *****
  • Posts: 2,130
  • Joined: 02 Mar 2019
  • Loc: Melbourne, Australia

Posted 24 October 2020 - 05:50 PM

This is a great comparison of the two techniques, I've been experimenting a bit with AstraImage but usually return to Registax.

 

What I find really interesting is the slight change in colour cast for these images, was this deliberate or just an artifact of the processing technique?

 

Thanks for posting, Andrew



#10 Jeff B1

Jeff B1

    Gemini

  • -----
  • Posts: 3,002
  • Joined: 07 Mar 2014
  • Loc: South Central Florida

Posted 24 October 2020 - 06:19 PM

Wow, the Mars #2 images has one of the most detailed Olympus Mons I've ever seen in an Earth-based image of Mars.  The caldera and general slope and scarf profile is like the Viking Orbiter images!  A 40-Km shadow computed just right for the Sun angle.

 

Once upon a time, way back in my first encounters in observing Mars with some huge telescopes and photographs of Mars for those times (40+ years ago) we only dreamed to see such images of the Red Planet.   Only HST and Mars orbiting machines showed us such clear and details of the surface of Mars.  Now it seems like some amateurs casually produce them at a whim.  cool.gif

 

Say, what kind of a telescope did you use, Hubble?  smile.gif   It must be cold up there in orbit......

 

Oops, a 14" Dob!!!!!  


Edited by Jeff B1, 24 October 2020 - 06:30 PM.

  • leoyasu likes this

#11 Jeff B1

Jeff B1

    Gemini

  • -----
  • Posts: 3,002
  • Joined: 07 Mar 2014
  • Loc: South Central Florida

Posted 24 October 2020 - 06:44 PM

In the past someone mused about, I paraphrase, “if only Einstein had a modern hand calculator in his younger life, how much more could he have achieved.”  Maybe if he had my laptop he could have resolved his General Relativity theory!   If one would use similar analogy in astronomical terms, we may compare the photographs and their new-age images of professional astronomers of the past, say just a few decades ago, to the electronic images of today’s amateur astronomer using modest equipment; such a deal, we may say.  How much more would we know if they had the same technology?


  • leoyasu likes this

#12 Yu Gu

Yu Gu

    Vostok 1

  • *****
  • Posts: 179
  • Joined: 18 Jun 2006

Posted 25 October 2020 - 08:01 AM

Thank you all for the posts and likes! waytogo.gif

 

For deconvolution I was experimenting with AstraImage, Fitswork and AstroSurface but I got more consistent results with AstraImage. The resulting image contained less noise, graininess and a bit more sharpness. I use the Lucy-Richardson deconvolution on the luminance channel. In particular, I apply between 10 and 30 iterations based on the data quality with a strength between 5 and 10. Then for the Blur kernel I use a size between 1.5 and 2. I'm pretty new to this, so I'm still gaining experience.

 

Leo.

Thanks for the helpful information! Since there are two factors at play here: 1.5x drizzle and deconvolution, would you mind do other experiment? i.e., compare the results of RS6 wavelet vs Astraimage deconvolution, both with 1.5x drizzled stack.

Clear skies,

Gu



#13 leoyasu

leoyasu

    Explorer 1

  • -----
  • topic starter
  • Posts: 64
  • Joined: 03 May 2013
  • Loc: Buenos Aires, Argentina

Posted 26 October 2020 - 05:57 PM

Thank you all!
 

This is a great comparison of the two techniques, I've been experimenting a bit with AstraImage but usually return to Registax.
 
What I find really interesting is the slight change in colour cast for these images, was this deliberate or just an artifact of the processing technique?
 
Thanks for posting, Andrew


Tulloch for the second run I changed the handling of colors a bit. Now I apply some saturation on both PS and RS6, then play a bit with the channel mixer.
 
 
 

Wow, the Mars #2 images has one of the most detailed Olympus Mons I've ever seen in an Earth-based image of Mars.  The caldera and general slope and scarf profile is like the Viking Orbiter images!  A 40-Km shadow computed just right for the Sun angle.
 
Once upon a time, way back in my first encounters in observing Mars with some huge telescopes and photographs of Mars for those times (40+ years ago) we only dreamed to see such images of the Red Planet.   Only HST and Mars orbiting machines showed us such clear and details of the surface of Mars.  Now it seems like some amateurs casually produce them at a whim.  cool.gif
 
Say, what kind of a telescope did you use, Hubble?  smile.gif   It must be cold up there in orbit......
 
Oops, a 14" Dob!!!!!

In the past someone mused about, I paraphrase, “if only Einstein had a modern hand calculator in his younger life, how much more could he have achieved.”  Maybe if he had my laptop he could have resolved his General Relativity theory!   If one would use similar analogy in astronomical terms, we may compare the photographs and their new-age images of professional astronomers of the past, say just a few decades ago, to the electronic images of today’s amateur astronomer using modest equipment; such a deal, we may say.  How much more would we know if they had the same technology?


Jeff I wonder about it constantly, thank you for the kind words! grin.gif

 

 

 

Thanks for the helpful information! Since there are two factors at play here: 1.5x drizzle and deconvolution, would you mind do other experiment? i.e., compare the results of RS6 wavelet vs Astraimage deconvolution, both with 1.5x drizzled stack.

Clear skies,

Gu

Of course! I made a comparison with the first Martian image. The only difference was sharpening, wavelets (my usual settings vs. decon). The rest of the post processing was the same in RS6 (saturation) and PS (levels, curves, exposure, smart sharpen, colors, etc.). In the version processed with wavelets the details seem to be less and with more artifacts. In addition, the blue channel seems displaced by the different focus it presents with respect to the rest of the channels. It seems that deconvolution reduces that:

 

2020-09-15-0545_1-LY-Mars_RGB_Wav.png

 

2020-09-15-0545_1-LY-Mars_RGB_Wav.gif

 

canales.png

 

Leo.


  • Yu Gu, dswtan, vilchez and 1 other like this

#14 Yu Gu

Yu Gu

    Vostok 1

  • *****
  • Posts: 179
  • Joined: 18 Jun 2006

Posted 26 October 2020 - 09:10 PM

Hi Leo,

That's really promising! I just downloaded AstraImage (still the trial version) and it does seem to have some advantages over RS6 on my images: finer details and less noise. I need to play with it more!

Gu



#15 raylinds

raylinds

    Explorer 1

  • *****
  • Posts: 96
  • Joined: 16 Sep 2020
  • Loc: Candlewood Lake, CT

Posted 27 October 2020 - 08:38 AM

Very impressive work. Thanks for sharing.



#16 Yu Gu

Yu Gu

    Vostok 1

  • *****
  • Posts: 179
  • Joined: 18 Jun 2006

Posted 27 October 2020 - 09:08 PM

Here is my version of comparison. The improvement is not as dramatic as Leo's examples but I am satisfied!

Gu

Attached Thumbnails

  • Deconvolution vs Wavelet.jpg

  • JMP and leoyasu like this


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