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

Jupiter processing tips requested (with free data!)

  • Please log in to reply
32 replies to this topic

#1 Tulloch

Tulloch

    Apollo

  • *****
  • topic starter
  • Posts: 1,362
  • Joined: 02 Mar 2019
  • Loc: Melbourne, Australia

Posted 29 May 2020 - 07:34 PM

Hi all, recently I got some good (but not great) data from Jupiter and some of its moons, and it's really the first time I've looked into processing data from my 9.25" SCT after upgrading from a 6". The larger OTA gives me the possibility of greater resolution but I'm having trouble with processing the data and balancing sharpness levels to produce an image I'm happy with.

 

I reckon that Jupiter is probably the hardest planet to process as it has a mixture of large and fine structures on its surface. Saturn is easier, the general rules I follow are; don't make the Cassini Division too large, sharpen to the point of showing surface details but don't make the planet look like a lollipop. The ice giants don't have much structure to show and Mars is still a bit of learning journey for me at this stage.

 

My problem is that I don't really know what I like, but I'm quite sure of what I don't like. I also don't have enough experience with working out how far I can push data to get the best result, so I'm throwing it out to you all to see what you can come up with. So, if you have the time and would like to play with some reasonable quality data, I'd be really interested to see what others can come up with.

 

The link below is a folder containing images of Jupiter and its moons was taken on 20200524 20:18:43 UT, from a 3 minute video with the best 5000, 7500 and 10000 frames stacked in AS!3 (from a total of 29801) using 3x drizzle. Jupiter was at an elevation of around 56* at the time of imaging. I thought the 7500 frame stack had the best mix of sharpness and noise and used it below, you may think differently.

 

https://drive.google...y60?usp=sharing

 

I've tried a number of approaches myself, and two I'm reasonably happy with (for just the planet area) are shown below, one sharpened in Registax and the other in Astra Image. The AI image used both deconvolution and wavelets, while Registax only uses wavelets. The Registax and Astra Image settings I used are also shown below, after these were applied I fixed the colour cast and levels in Photoshop with a bit of added noise reduction (and Gaussian blur in the AI image) to taste. However to my eye, the resolution in the Registax processing looks too low, while the AI image looks over sharpened. However, maybe this is just an artifact of this data set and it's not possible to do better, however I'm not experienced enough to know better.

 

So I'm opening it up to anyone who would like to have a play with this data and see what you can come up with. If you do have a go I'd be very interested in your process so I might be able to incorporate it into my own, or at least it would give me some tips on which way to go.

 

Thanks, Andrew

 

Attached Thumbnails

  • Jupiter AstraImage vs Registax.png
  • Registax settings for comparison.PNG
  • Astra Image settings for comparison.PNG

  • Kenny V., John Rogers, troyt and 2 others like this

#2 sunnyday

sunnyday

    Cosmos

  • *****
  • Posts: 7,878
  • Joined: 23 Sep 2019
  • Loc: Ottawa,Canada

Posted 29 May 2020 - 07:48 PM

very beautiful pictures .
I prefer the one with registax.
I find it in more detail, also I find the larger photo on the left, may be a bivision effect.

thanks 


  • Tulloch likes this

#3 Kokatha man

Kokatha man

    Voyager 1

  • *****
  • Posts: 14,005
  • Joined: 13 Sep 2009
  • Loc: "cooker-ta man" downunda...

Posted 29 May 2020 - 08:23 PM

<"My problem is that I don't really know what I like, but I'm quite sure of what I don't like."> Probably it's best worth working on that particular aspect further Andrew - it's what most people do iirc! :lol:

 

I've always found old Reggie's wavelets the best, from fine detail with the first sliders to etc - I have several other softwares to choose from but he still comes up ace for me. waytogo.gif

 

I found out that I could simply enter my old license key to get the latest AstraImage version after someone informed me of this fact - & AI replied to an email I sent them...so I'll play around & familiarise myself with all the new bells & whistles...but I doubt that much of it will display much difference... ;)


  • Tulloch likes this

#4 Tulloch

Tulloch

    Apollo

  • *****
  • topic starter
  • Posts: 1,362
  • Joined: 02 Mar 2019
  • Loc: Melbourne, Australia

Posted 29 May 2020 - 08:39 PM

<"My problem is that I don't really know what I like, but I'm quite sure of what I don't like."> Probably it's best worth working on that particular aspect further Andrew - it's what most people do iirc! lol.gif

 

I've always found old Reggie's wavelets the best, from fine detail with the first sliders to etc - I have several other softwares to choose from but he still comes up ace for me. waytogo.gif

 

I found out that I could simply enter my old license key to get the latest AstraImage version after someone informed me of this fact - & AI replied to an email I sent them...so I'll play around & familiarise myself with all the new bells & whistles...but I doubt that much of it will display much difference... wink.gif

Thanks Darryl, I'm working on finding a style I prefer - new equipment brings new expectations and requires different processing.

 

I've found that the deconvolution step in AI can reduce the blur significantly without introducing a lot of noise, as long as you get the "kurtosis" value correct - it's extremely sensitive to slight changes in value. 


Edited by Tulloch, 29 May 2020 - 08:40 PM.


#5 AstroEthan

AstroEthan

    Vanguard

  • -----
  • Posts: 2,173
  • Joined: 16 Jun 2014
  • Loc: San Antonio, Texas, United States

Posted 29 May 2020 - 08:56 PM

First Darren shares his data, now you too?

 

Guess I'll play with your set once I'm caught up with mine. Personally, I prefer your RegiStax result.

 

EDIT: Never seen artifacts like this. Are your PNGs compressed in a lossy fashion?

 

2020-05-24-2020_2-L-Jup_AS_F7500_l6_ap128_Driz30.png


Edited by AstroEthan, 29 May 2020 - 08:58 PM.


#6 Tulloch

Tulloch

    Apollo

  • *****
  • topic starter
  • Posts: 1,362
  • Joined: 02 Mar 2019
  • Loc: Melbourne, Australia

Posted 29 May 2020 - 09:29 PM

First Darren shares his data, now you too?

 

Guess I'll play with your set once I'm caught up with mine. Personally, I prefer your RegiStax result.

 

EDIT: Never seen artifacts like this. Are your PNGs compressed in a lossy fashion?

 

attachicon.gif2020-05-24-2020_2-L-Jup_AS_F7500_l6_ap128_Driz30.png

Well, Darren first shared his data a year ago now (May 31, 2019), mine is nowhere near as good, but why not? 

https://www.cloudyni...to-a-good-home/

 

I can't see those artifacts that you can, these were directly out of AS!3 in 48bit uncompressed png format. Are you sure you don't have oil on your screen lol.gif ?

 

Thanks, Andrew



#7 Sunspot

Sunspot

    Cosmos

  • *****
  • Posts: 9,337
  • Joined: 15 Mar 2005
  • Loc: Surprise, AZ

Posted 29 May 2020 - 10:30 PM

I don't see those artifacts either, odd. I've never had any success with drizzled data in the past. Maybe I'll give it a whirl.



#8 AstroEthan

AstroEthan

    Vanguard

  • -----
  • Posts: 2,173
  • Joined: 16 Jun 2014
  • Loc: San Antonio, Texas, United States

Posted 29 May 2020 - 11:54 PM

Figured out what was going on. PixInsight only handles 8-bit PNGs as far as I know (certainly only saves them as such). Converting them to TIFFs fixed it.

 

Wasn't aware of Darren's data share a year ago. They were recently posted on AstroBin, so I'll take a crack at them eventually. All I've done so far is run the recordings through DeTeCt and it found no impacts.


  • Tulloch likes this

#9 KiwiRay

KiwiRay

    Viking 1

  • -----
  • Posts: 738
  • Joined: 09 Apr 2015
  • Loc: Seattle, USA

Posted 30 May 2020 - 12:48 AM

I used some pre-saved Registax settings for this, and then brightened it up a bit in Photoshop.  Nice data.

 

2020-05-24-2020_2-L-Jup_AS_F10000_l6_ap128_Driz30_Rb.png

 

There are many ways to achieve similar results, but my RS settings only used the first three sliders at 100, 35.5 and 10.4% respectively. Denoise settings were 0.65, 0.6, 0.55, and Sharpen were 0.25, 0.17 and 0.15; linear, Gaussian, with Linked Wavelets checked.

 

Thanks for sharing.  I have nothing of my own to play with, so that was fun.


Edited by KiwiRay, 30 May 2020 - 02:50 PM.

  • Kenny V., Tulloch and sunnyday like this

#10 Tulloch

Tulloch

    Apollo

  • *****
  • topic starter
  • Posts: 1,362
  • Joined: 02 Mar 2019
  • Loc: Melbourne, Australia

Posted 30 May 2020 - 01:23 AM

Thanks Ray (I assume), that exactly the sort of thing I was interested in, something completely different from my own experience with Registax. Sharp enough, but refined to not look overblown and have an overly thick brown look which is what I normally get from Registax. I've never been a fan of linked wavelets before as the image always came out too harsh for me, but you've really got a good mix there.

 

Thanks again,

 

Andrew



#11 Tulloch

Tulloch

    Apollo

  • *****
  • topic starter
  • Posts: 1,362
  • Joined: 02 Mar 2019
  • Loc: Melbourne, Australia

Posted 30 May 2020 - 01:27 AM

Figured out what was going on. PixInsight only handles 8-bit PNGs as far as I know (certainly only saves them as such). Converting them to TIFFs fixed it.

 

Wasn't aware of Darren's data share a year ago. They were recently posted on AstroBin, so I'll take a crack at them eventually. All I've done so far is run the recordings through DeTeCt and it found no impacts.

I had a crack at Darren's data when he first posted it here, and all I remember was that it was almost impossible to get a final result that didn't look amazing :)

 

Unfortunately it spoiled me for my own meagre offerings...



#12 KiwiRay

KiwiRay

    Viking 1

  • -----
  • Posts: 738
  • Joined: 09 Apr 2015
  • Loc: Seattle, USA

Posted 30 May 2020 - 09:19 AM

I've never been a fan of linked wavelets before as the image always came out too harsh for me, but you've really got a good mix there.

 

If you checked Linked Wavelets with your settings above, it would look a mess.  It's just a different way of getting the same thing - go softer on the sliders (except the first) and the sharpening, and Linked Wavelets can work well.
 



#13 lgwong

lgwong

    Explorer 1

  • -----
  • Posts: 97
  • Joined: 13 Aug 2016

Posted 30 May 2020 - 11:18 AM

My version of your data, see whether you like it.

2020-05-24-2020_2-L-Jup_AS_F10000_l6_ap1


  • dhammy and sunnyday like this

#14 dhammy

dhammy

    Viking 1

  • -----
  • Posts: 728
  • Joined: 20 Jul 2015
  • Loc: Puerto Rico

Posted 30 May 2020 - 12:25 PM

Yeah the data is all there - now you want to capture a bunch of images of this quality to derotate - that will make the final image much better. 

 

I like Igwong's processing because it doesn't look like Jupiter has been pushed too hard in the processing. Have an image of Jupiter taken by the Hubble open beside you as you process it. The more you push the wavelets 4, 5, 6 the more unreal or over processed it tends to look to me personally. I always start with the 1 then move to the 2 and sometimes use the 3. 

 

So this is a quick one I did at a slightly larger scale which makes it less sharp but anyway. I used the 10,000 stack. I've included the wavelet settings, I did the auto RGB align, RGB Balance and resized to 40% using Bell. In Photoshop I adjusted the levels, the vibrance and the tone of the image. Further playing with these might give some sharper results. 

Attached Thumbnails

  • Jupiter test.jpg
  • Wavelet settings.JPG

Edited by dhammy, 30 May 2020 - 12:29 PM.

  • sunnyday likes this

#15 Toddeo

Toddeo

    Apollo

  • *****
  • Posts: 1,298
  • Joined: 19 Jan 2012
  • Loc: Sierra Vista, AZ

Posted 30 May 2020 - 12:48 PM

Here's my take. Just using Registax6. Can't seem to get rid of the "grid pattern".

Attached Thumbnails

  • CloudyNights #2.jpg
  • CloudyNight #4 ASF500.jpg

Edited by Toddeo, 30 May 2020 - 03:52 PM.

  • KiwiRay and sunnyday like this

#16 sunnyday

sunnyday

    Cosmos

  • *****
  • Posts: 7,878
  • Joined: 23 Sep 2019
  • Loc: Ottawa,Canada

Posted 30 May 2020 - 12:52 PM

wow, this is beautiful.
nice work guys
thanks.



#17 Tulloch

Tulloch

    Apollo

  • *****
  • topic starter
  • Posts: 1,362
  • Joined: 02 Mar 2019
  • Loc: Melbourne, Australia

Posted 30 May 2020 - 05:48 PM

Thanks to everyone who had a go processing this data, there is certainly a good range of different approaches for the same dataset, from the hard focus on details to a more "dreamy" appearance when the clouds softly merge into one another. Since moving up from the 6" where the OTA simply wasn't capable of producing this much detail, I guess I've been too focused on bring up as much resolution as I could get. This exercise has certainly shown me that just because you can do something, doesn't necessarily mean you should.

 

My version of your data, see whether you like it.

Thanks Igwong, you have certainly nailed the dream-like appearance, soft and beautiful.

 

Yeah the data is all there - now you want to capture a bunch of images of this quality to derotate - that will make the final image much better. 

 

I like Igwong's processing because it doesn't look like Jupiter has been pushed too hard in the processing. Have an image of Jupiter taken by the Hubble open beside you as you process it. The more you push the wavelets 4, 5, 6 the more unreal or over processed it tends to look to me personally. I always start with the 1 then move to the 2 and sometimes use the 3. 

 

So this is a quick one I did at a slightly larger scale which makes it less sharp but anyway. I used the 10,000 stack. I've included the wavelet settings, I did the auto RGB align, RGB Balance and resized to 40% using Bell. In Photoshop I adjusted the levels, the vibrance and the tone of the image. Further playing with these might give some sharper results. 

Thanks David, you seem to have taken Igwong's dreamlike appearance and upped it a bit :). Personally I don't like Hubble's rendition of the planets, I can't get over the colour cast it applies due to the filters it uses which are tuned for DSO. I find Cassini's images are closer to reality, but they are pretty old now. However, I get your point about having a point of reference to ground my own data.

 

This was one of the images from the set I took recently showing the appearance of Ganymede (taken at the same time as Darren took his), so I have data that could be used to derotate and stack, however WJ doesn't like moons (or their shadows) in the image, and I'd have to remove the field rotation manually and I always get that wrong which ends up making the final image look 

 

Here's my take. Just using Registax6. Can't seem to get rid of the "grid pattern".

Thanks Toddeo, I like the level of detail still present in the data without being overblown. To remove the "grid pattern", I generally do a sub-pixel blur in Photoshop which usually does the trick.

 

Thanks again all, lots to think about and play around with. I appreciate the time you have put on processing this data.

 

Andrew 

worse.



#18 RBChris

RBChris

    Messenger

  • -----
  • Posts: 406
  • Joined: 18 Nov 2013
  • Loc: Northern California

Posted 31 May 2020 - 05:03 PM

Hi Andrew,

 

Sorry I'm late to this topic - just saw the post. You have good data and you and the other posters have done a fine job with it. For what it's worth, here's what I did with PixInsight. It was done primarily with deconvolution (Restoration Filter) with a touch of wavelets. 

 

Randy

Attached Thumbnails

  • 2020-05-24-2020_2-L-Jup_AS_F7500_l6_ap128_Driz30.v2_small.jpg

  • Tulloch likes this

#19 Kokatha man

Kokatha man

    Voyager 1

  • *****
  • Posts: 14,005
  • Joined: 13 Sep 2009
  • Loc: "cooker-ta man" downunda...

Posted 31 May 2020 - 09:03 PM

...I must be struggling to find something to do Andrew - can't even go for my morning walk because the rain is teeming down! lol.gif

 

That all in the context of I never do processings of other people's work - for no reason other than it is their work & they have to find a balance that suits them...& processing someone else's work does neither person any real favours cos the original imager will probably think "nope not really my cup of tea!" & the re-processor thinks "I would've probably captured differently!" rofl2.gif

 

But fwiw here's a quick take from my perspective...didn't give much thought to the moons & it was quick & it isn't my data...similar in ways to other re-pros above probably...& left a wee bit of noise in 'cos it seemed a tad better doing so. wink.gif

 

EDIT: probably a bit brighter helps - but it's still mainly "in the eye of the beholder..!" lol.gif Also, for simplicity the applications were pretty much those I demonstrate in my processing tutes on our website... ;)

 

2020-05-24-2020_2-L-Jup_AS_F5000@100%WEB-lvls.png


Edited by Kokatha man, 31 May 2020 - 10:18 PM.

  • RBChris and Tulloch like this

#20 Tulloch

Tulloch

    Apollo

  • *****
  • topic starter
  • Posts: 1,362
  • Joined: 02 Mar 2019
  • Loc: Melbourne, Australia

Posted 01 June 2020 - 12:16 AM

Thanks Randy and Darryl, I find it very interesting that everyone has their own interpretation of the same dataset leading to quite different final results. 

 

I think that one of the hardest things to learn is how to process the data, and since there's no "Standard" data to work from, it's hard to know whether what you are doing is "correct", or "best" or even "adequate" since everyone's raw data is so different. For instance, it might answer my question; "Is it my processing skills that are inadequate, or is it just that my data is so poor it's impossible to do any better?"

 

It might be good if there was a few standard datasets available for newbies/novices (including myself) to practice our processing skills on and publish the results, (eg data from a 6", 10", 14" in a variety of seeing conditions) the main downside to that might be "groupthink" where everyone ends up processing the data the same way and variations are discouraged. 

 

There are of course some excellent tutorials online which discuss Registax (notably Steve's tutorial videos and Darryl's step-by-step web tutorials) and other processing tools, but the settings you need to use are very dependent on the size and quality of your own original data, so it's difficult to "play along at home".

 

Anyway, I really appreciate the time and effort from everyone (even if it was just because it's pouring with rain outside and you are stuck at home with nothing to do lol.gif), it has certainly given me (and I hope others) something to think about with my own processing in the future flowerred.gif.

 

Andrew


  • RBChris likes this

#21 RedLionNJ

RedLionNJ

    Soyuz

  • *****
  • Posts: 3,844
  • Joined: 29 Dec 2009
  • Loc: Red Lion, NJ, USA

Posted 03 June 2020 - 02:30 PM

Sorry, I'm a bit late to the game, fellows. Been away from imaging for a while.

 

Here's a purely Registax solution:

 

l2.jpg

 

This was from the 7500 stack. Noise is still pretty high, though.

 

 

Grant


  • Jeffrey C. and Tulloch like this

#22 DMach

DMach

    Apollo

  • -----
  • Posts: 1,298
  • Joined: 21 Nov 2017
  • Loc: The most light-polluted country in the world :(

Posted 10 June 2020 - 04:46 AM

Hi Andrew, here's my attempt. Sorry it took me so long to get around to it, things have been a bit hectic lately.

 

Tried to strike a balance between detail and softness ... possibly I pushed the sharpening a little far (#4 in particular).

 

Andrew_2020-05-24-2020_2-L-Jup_AS_F5000_l6_ap128_Driz30_30pc_ba.png

 

A little bit of a rush job as the kids are in my ear ... and I just realised I forgot to take a screen grab of the wavelet settings! But they were something like this:

 

Andrew_wavelets.png

 

I've been playing with setting the higher wavelets (5 and 6) less than one lately ... this selectively adds some smoothing to the larger structures and allows you to push wavelets 3 and 4 a bit further than you otherwise might.

 

The moons were processed with a different wavelet setting, as I find some sharpening at levels 5 & 6 help mitigate the "black hole" effect of sharpening the moons ... if that description makes sense lol.

 

A small amount of additional sharpening was done in Astra Image PLUS ... this comes from Darryl's workflow, from memory. In Photoshop, I blend this deconvolution boost with the image straight out of Registax to taste and then adjust denoise, do a little colour boost and play with gamma/exposure.

 

I prefer to manage denoise in Photoshop these days rather than applying it in Registax, for Jupiter at least (which does have fine details and structure). Saturn is a different beast lol.



#23 Tulloch

Tulloch

    Apollo

  • *****
  • topic starter
  • Posts: 1,362
  • Joined: 02 Mar 2019
  • Loc: Melbourne, Australia

Posted 10 June 2020 - 07:58 AM

Thanks Darren, that's a really nice result from a seriously different set of wavelets! No de-noise at all, I use quite high numbers myself in Registax, looks like I'll need to play in Photoshop a bit more.

 

It's going to take me a while to go through it properly but it certainly has your trademark "look and feel" for Jupiter, really attractive processing.

 

Thanks again,

 

Andrew



#24 BQ Octantis

BQ Octantis

    Mercury-Atlas

  • *****
  • Posts: 2,761
  • Joined: 29 Apr 2017
  • Loc: Red Centre, Oz

Posted 11 June 2020 - 02:45 AM

Yay!…free data!

 

I once heard, "If something is free, then the product is probably you." laugh.gif
 

(Click for full size.)

post_BQ.jpg

 

Looks even better on the phone…is it worthy of tips?


Cheers,

 

BQ


Edited by BQ Octantis, 11 June 2020 - 04:24 AM.

  • Foc and Tulloch like this

#25 treelo

treelo

    Sputnik

  • -----
  • Posts: 30
  • Joined: 12 Nov 2018

Posted 11 June 2020 - 04:26 AM

these are amazing




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