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

Improved Jupiter and Processing Workflow

  • Please log in to reply
9 replies to this topic

#1 aeroman4907

aeroman4907

    Viking 1

  • -----
  • topic starter
  • Posts: 957
  • Joined: 23 Nov 2017
  • Loc: Castle Rock, Colorado

Posted 12 July 2019 - 05:59 PM

As I recently commented regarding deconvolution of images with PI using greyscale derived from a synthetic luminance channel, I decided to take things a bit further with testing on a small image size of Jupiter.  I used AS!3 to save a .fit file and work in 32 bit.  I then deconvolved and sharpened synthetic R, G, B, and L files in PI, and then recombined.  I then saved as a 16 bit .tiff and did very minor final adjustments in PS.  The data was not great to begin with as I described in a prior post, but I think there are definitely some improvements.  New image is on the left with a bit greater resolution and more enhanced color range (saturation is about the same).  The Jupiter on the left is actually slightly larger as the older deconvolution method actually started to reduce the outer limb of Jupiter.

 

I am now interested to see how this work on my lunar images, but I don't have any original video files, so I can't create a 32 bit .fit file until I capture new video.  Might not be worth the extra file size on the moon, but I did notice the noise levels were lower as well working with the Jupiter images.

 

Jupiter-Comparison.jpg


  • Sunspot, mdowns, Billytk and 1 other like this

#2 petert913

petert913

    Mercury-Atlas

  • *****
  • Posts: 2916
  • Joined: 27 May 2013
  • Loc: Portland, OR

Posted 12 July 2019 - 06:18 PM

The colors are outstanding


  • aeroman4907 likes this

#3 Kokatha man

Kokatha man

    Voyager 1

  • *****
  • Posts: 12610
  • Joined: 13 Sep 2009
  • Loc: "cooker-ta man" downunda...

Posted 12 July 2019 - 06:36 PM

Hi bro - I don't like to be the bearer of bad news, but the image on the right has (very) slightly better resolution/detail. ;)

 

But keep up the experimentation - it's the only way any of us progress! waytogo.gif



#4 aeroman4907

aeroman4907

    Viking 1

  • -----
  • topic starter
  • Posts: 957
  • Joined: 23 Nov 2017
  • Loc: Castle Rock, Colorado

Posted 13 July 2019 - 07:03 AM

Hi bro - I don't like to be the bearer of bad news, but the image on the right has (very) slightly better resolution/detail. wink.gif

 

But keep up the experimentation - it's the only way any of us progress! waytogo.gif

 

I think this particular image posted a bit funny, so I'll repost at a slightly bigger size at original capture size.  If you look at what is visible in the equatorial bands, I definitely observe items better in the new processing (at the top) than those at the bottom.  I wouldn't say that the new image has a ton more resolution because it doesn't, but it has definitely more color and details visible.  I certainly wouldn't say the old process has more resolution.

 

Regardless, thanks for your compliment regarding experimentation.  cool.gif 

 

Band-Comparison.jpg


  • Peregrinatum likes this

#5 Peregrinatum

Peregrinatum

    Ranger 4

  • *****
  • Posts: 307
  • Joined: 27 Dec 2018

Posted 13 July 2019 - 03:57 PM

As I recently commented regarding deconvolution of images with PI using greyscale derived from a synthetic luminance channel, I decided to take things a bit further with testing on a small image size of Jupiter.  I used AS!3 to save a .fit file and work in 32 bit.  I then deconvolved and sharpened synthetic R, G, B, and L files in PI, and then recombined.  I then saved as a 16 bit .tiff and did very minor final adjustments in PS.  The data was not great to begin with as I described in a prior post, but I think there are definitely some improvements.  New image is on the left with a bit greater resolution and more enhanced color range (saturation is about the same).  The Jupiter on the left is actually slightly larger as the older deconvolution method actually started to reduce the outer limb of Jupiter.

 

I am now interested to see how this work on my lunar images, but I don't have any original video files, so I can't create a 32 bit .fit file until I capture new video.  Might not be worth the extra file size on the moon, but I did notice the noise levels were lower as well working with the Jupiter images.

 

attachicon.gif Jupiter-Comparison.jpg

I really like this approach and it is one that I have played around with some but it hasn't born much fruit just yet... I agree that the channel separated image looks to have more detail... I am curious what process do you like to use for sharpening in PI?


  • aeroman4907 likes this

#6 Lunatiki

Lunatiki

    Vanguard

  • *****
  • Posts: 2266
  • Joined: 07 Aug 2005
  • Loc: Amarillo, TX

Posted 13 July 2019 - 04:29 PM

Your first sentence flew 36,000 feet right over my head. Outstanding picture, whatever you did. When it comes to imaging, I'm kind of like a race car driver. I can really drive well, but when it comes to the mechanics of the car, I'm a challenged. I'm hoping to change that this go around.


  • aeroman4907 likes this

#7 Tulloch

Tulloch

    Mariner 2

  • -----
  • Posts: 226
  • Joined: 02 Mar 2019
  • Loc: Melbourne, Australia

Posted 13 July 2019 - 05:16 PM

I think this particular image posted a bit funny, so I'll repost at a slightly bigger size at original capture size.  If you look at what is visible in the equatorial bands, I definitely observe items better in the new processing (at the top) than those at the bottom.  I wouldn't say that the new image has a ton more resolution because it doesn't, but it has definitely more color and details visible.  I certainly wouldn't say the old process has more resolution.

 

Regardless, thanks for your compliment regarding experimentation.  cool.gif

 

attachicon.gif Band-Comparison.jpg

Hi there, the second image certainly shows the difference better. 

 

I'm no expert in this area, but in my limited experience it appears that with a bit of sharpening and saturation, the two images might look identical. I hope you don't mind, but I grabbed your second image and did a quick 10 minute test to remove some Gaussian blur and upping the saturation in Photoshop Elements and got this result. You should be able to do much better with the original data (if you want to of course smile.gif ).

 

I don't mean this as a criticism of your new technique, just wondering if all the additional work to deconvolve the colours is worth it? Maybe there are other advantages that weren't immediately obvious to me.

 

Thanks, Andrew

Attached Thumbnails

  • post-285348-0-53772800-1563019338 edit2.jpg

  • aeroman4907 likes this

#8 aeroman4907

aeroman4907

    Viking 1

  • -----
  • topic starter
  • Posts: 957
  • Joined: 23 Nov 2017
  • Loc: Castle Rock, Colorado

Posted 14 July 2019 - 08:05 AM

Hi there, the second image certainly shows the difference better. 

 

I'm no expert in this area, but in my limited experience it appears that with a bit of sharpening and saturation, the two images might look identical. I hope you don't mind, but I grabbed your second image and did a quick 10 minute test to remove some Gaussian blur and upping the saturation in Photoshop Elements and got this result. You should be able to do much better with the original data (if you want to of course smile.gif ).

 

I don't mean this as a criticism of your new technique, just wondering if all the additional work to deconvolve the colours is worth it? Maybe there are other advantages that weren't immediately obvious to me.

 

Thanks, Andrew

I don't mind Andrew.  I personally see two things with your reprocessing the lower half.  One is that I think the image is being pushed a little too far and the contrasts now look over-processed and has less of a 'photo' look.  I could also push the upper image harder, but I feel that is already at the limit of my personal tastes.  Secondly, the bottom image has increased the saturation, but of fewer tonal values.  I think part of the difference in these color conversations are likely due to people's monitor's and ability to see color.  I use a Asus PA328Q monitor (recently replaced by the PA329Q) which has similar specs.  The color gamut is very wide and highly accurate.  I have noticed before that I can clearly see issues with other people's images (particularly with black or shadow levels) that others cannot see on their monitor without taking measurements.  So even with the adjustments you made, the winner in my book is still very clearly the upper image.

 

To further illustrate, I have attached a hyper saturated version comparing the two images.  Both are on the same combined image, so any adjustments are identical.  Of course the image is garish now, but the variation in color is definitely more prevalent on the left hand image.

 

Extreme.jpg

 

With regard to the extra work to deconvolve the colors, it is pretty minimal for my workflow.  I am used to working on whole moon mosaics, so processing takes many hours to complete.  Separate deconvolution of the three color channels and a few other various steps takes about 5 minutes at most, so some slight color improvement with that timeframe is easily acceptable to me.

 

Thanks for your thoughts.


Edited by aeroman4907, 14 July 2019 - 08:12 AM.

  • Tulloch likes this

#9 aeroman4907

aeroman4907

    Viking 1

  • -----
  • topic starter
  • Posts: 957
  • Joined: 23 Nov 2017
  • Loc: Castle Rock, Colorado

Posted 14 July 2019 - 08:23 AM

I really like this approach and it is one that I have played around with some but it hasn't born much fruit just yet... I agree that the channel separated image looks to have more detail... I am curious what process do you like to use for sharpening in PI?

Thanks.  Regarding the deconvolution process, I first run Restoration Decon, and then I perform a second run with the Decon process.  That provides most of the sharpening.  I do use a little Atrous wavelet sharpening as well, but that is very light.  There is also a couple of runs of TGV noise reduction because I really don't like noise in my images, particularly with planets.

 

Each image and target (i.e. Jupiter, Moon, Saturn) etc. is pretty unique to processing, so providing any values I used would be completely useless to give to you.  All I would say is experiment heavily with these features in PI and you'll come away with a workflow that works for you.  I have spent many hundreds of hours deconvolving lunar images, so I have a natural feel for how the program works and I simply needed to make some adjustments to how the process works on an image like Jupiter with the contrast levels and details much less defined than the moon.  As you can tell, I am still learning what works best.



#10 Tulloch

Tulloch

    Mariner 2

  • -----
  • Posts: 226
  • Joined: 02 Mar 2019
  • Loc: Melbourne, Australia

Posted 14 July 2019 - 09:24 AM

I don't mind Andrew.  I personally see two things with your reprocessing the lower half.  One is that I think the image is being pushed a little too far and the contrasts now look over-processed and has less of a 'photo' look.  I could also push the upper image harder, but I feel that is already at the limit of my personal tastes.  Secondly, the bottom image has increased the saturation, but of fewer tonal values.  I think part of the difference in these color conversations are likely due to people's monitor's and ability to see color.  I use a Asus PA328Q monitor (recently replaced by the PA329Q) which has similar specs.  The color gamut is very wide and highly accurate.  I have noticed before that I can clearly see issues with other people's images (particularly with black or shadow levels) that others cannot see on their monitor without taking measurements.  So even with the adjustments you made, the winner in my book is still very clearly the upper image.

 

To further illustrate, I have attached a hyper saturated version comparing the two images.  Both are on the same combined image, so any adjustments are identical.  Of course the image is garish now, but the variation in color is definitely more prevalent on the left hand image.

 

attachicon.gif Extreme.jpg

 

With regard to the extra work to deconvolve the colors, it is pretty minimal for my workflow.  I am used to working on whole moon mosaics, so processing takes many hours to complete.  Separate deconvolution of the three color channels and a few other various steps takes about 5 minutes at most, so some slight color improvement with that timeframe is easily acceptable to me.

 

Thanks for your thoughts.

No worries, I don't usually like processing other people's images so I'm glad you didn't mind (at least not too much smile.gif ). With regards to monitors and colour matching, I have found this particularly difficult since I use a laptop for processing, and laptop screens are notoriously variable with aspect angle. I recently colour calibrated my laptop screen with one of those USB spyder things, so my colours look more accurate now, but I know that using a laptop screen is not ideal at all.

 

I do tend to over-sharpen my images, so if you think it was too much then that's fine, they are your images! It sounded to me like a lot more processing but if it's just a matter of 5 minutes for the new processing technique then that's not an issue.

 

Andrew




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