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

Using the quiet period to work on image processing

  • Please log in to reply
22 replies to this topic

#1 jwestervelt

jwestervelt

    Viking 1

  • -----
  • topic starter
  • Posts: 554
  • Joined: 01 Sep 2012
  • Loc: Phoenix, Arizona

Posted 20 April 2019 - 05:57 PM

I think I figured out how to do that "3D" effect...

I'm still working out a solid method, but I have found a few guides out there and have adapted them to my current software setup.

This is the result of revisiting my capture from 13 April.

get.jpg?insecure


 


  • bill w, TerryWood, Great Attractor and 2 others like this

#2 descott12

descott12

    Viking 1

  • -----
  • Posts: 867
  • Joined: 28 Sep 2018
  • Loc: Charlotte, NC

Posted 20 April 2019 - 06:56 PM

That looks nice but the image is inverted. As you can see the umbra and limb are bright and should be dark.

Are you using the "difference" merge mode between layers?  I have been teaching myself the same stuff recently and this was very similar to my first results (using Gimp and the difference mode). Of course, inverted/non-inverted is probably more a personal preference than anything else.



#3 jwestervelt

jwestervelt

    Viking 1

  • -----
  • topic starter
  • Posts: 554
  • Joined: 01 Sep 2012
  • Loc: Phoenix, Arizona

Posted 20 April 2019 - 07:14 PM

It is inverted, but that's ultimately what gives it the 3D type of effect where the lighter colored areas seem to be lifted from the surface.  I was noticing this in other shots, most notably Maurits' stuff.

The curves are only partially inverted, which is why the proms are still light and space is still black.  I found a guide that went over the process and I am still learning a fair amount through trial and error.


For what it is worth, this is a single layer.  It was a single stack all taken at the same exposure level.   


Edited by jwestervelt, 20 April 2019 - 07:15 PM.


#4 descott12

descott12

    Viking 1

  • -----
  • Posts: 867
  • Joined: 28 Sep 2018
  • Loc: Charlotte, NC

Posted 20 April 2019 - 07:18 PM

That is really great that you got proms without blowing out the surface. What exposure/gain did you use? Were you simply able to stretch the prom colors without destroying the surface? I think we have almost the same setup so I'd like to try that.



#5 rigel123

rigel123

    Hubble

  • *****
  • Posts: 16356
  • Joined: 29 Jun 2009
  • Loc: SW Ohio

Posted 20 April 2019 - 07:41 PM

Looks nice!  Do you use Photoshop? If so you could process it twice, once for proms and once for the surface, then combine them and use the Free Transform to enlarge the surface image to overlay that bright ring you have around the limb.  With the surface inverted it makes it fairly easy to blend in with the prom layer.



#6 MalVeauX

MalVeauX

    Fly Me to the Moon

  • *****
  • Posts: 5294
  • Joined: 25 Feb 2016
  • Loc: Florida

Posted 20 April 2019 - 09:29 PM

Heya,

 

I find that the key to a more 3D like sphere or limb shot is really the presence of a gradient from the disc to the limb, whether its inverted or not. The flatter the disc, the less 3D it seems to look. You can test this by taking the same disc and processing it to be flat from center to edge and compare that to one with a gradient towards the limb.

 

And of course, inverted & falsely colored is a personal preference. Some like it. Some loath it.

 

Very best,


  • rigel123, jwestervelt and torsinadoc like this

#7 jwestervelt

jwestervelt

    Viking 1

  • -----
  • topic starter
  • Posts: 554
  • Joined: 01 Sep 2012
  • Loc: Phoenix, Arizona

Posted 20 April 2019 - 11:44 PM

That is really great that you got proms without blowing out the surface. What exposure/gain did you use? Were you simply able to stretch the prom colors without destroying the surface? I think we have almost the same setup so I'd like to try that.

For this shot, I had gain at zero and exposure at 2.2ms, 1x1 binning, "16" bit capture (actually only 14).  Lately I've been setting gain to 25 and adjusting exposure until the upper end of the histogram sits at about 85-90.

Basic image acquisition process is rough focus, rotate camera to align field to celestial pole, fine focus.  Then I turn on the histogram for SharpCap and pressure tune the scope and try to get the median of the histogram as low as far to the right as possible.  At this point, a majority of the photosphere is being blocked, but it may be necessary to fine-tune it to a certain feature.  I don't think a single-stack is selective enough to make doppler tuning a thing yet.  I'll find out when I can doublestack without having my wife murder me.

After tuning, I adjust the exposure to get it so the upper end of the histogram is around 85-90% of full saturation.  I doublecheck the offset to make sure the black value is as low as I can get it.

 

I grab 1000 images, which corresponds to about 20 seconds of time, probably adequate to avoid motion blur due to an active sun. 
I then stack the images in autostakkert 3.0, and apply the 1.5x drizzle enhancement.  3.0x is too much and there isn't much improvement over 1.5x, but 1.5x does help knock down the grid pattern a bit from the imx178 sensor based cameras while bringing out a bit of subpixel info.

After that, I take the image into imppg and gimp.  I have a video covering what I do in imppg, it is a bit much to describe in text.  I'll link that as soon as youtube finishes processing the video.
 

Looks nice!  Do you use Photoshop? If so you could process it twice, once for proms and once for the surface, then combine them and use the Free Transform to enlarge the surface image to overlay that bright ring you have around the limb.  With the surface inverted it makes it fairly easy to blend in with the prom layer.

No photoshop.  I only used Gimp to flip image horizontally and do final level adjustments. This is only a single image, there isn't a prom layer and a disk layer being stacked.  I'm using a trick inside of imppg to pull this off, which is why there is a bright limb.  It is important to note that this is a compromise due to not having a double-stacked scope.  The second etalon will set me back $850 and some change.  My wife is OK with $500... she starts eyeing the larger kitchen knives when I mention $600... and she grabs one of them if I start talking about $750 and up.  lol.gif

I am writing up a start-to-finish guide, mainly for my wife so she can use the scope when I'm not around and get her own images. I'll hopefully be able to share this soon.
 


Edited by jwestervelt, 20 April 2019 - 11:50 PM.

  • torsinadoc likes this

#8 jwestervelt

jwestervelt

    Viking 1

  • -----
  • topic starter
  • Posts: 554
  • Joined: 01 Sep 2012
  • Loc: Phoenix, Arizona

Posted 21 April 2019 - 12:32 AM

Video covering my ImPPG process:

 

https://www.youtube....h?v=0LpiVFcabbo


  • torsinadoc and MalVeauX like this

#9 rigel123

rigel123

    Hubble

  • *****
  • Posts: 16356
  • Joined: 29 Jun 2009
  • Loc: SW Ohio

Posted 21 April 2019 - 06:45 AM

For this shot, I had gain at zero and exposure at 2.2ms, 1x1 binning, "16" bit capture (actually only 14).  Lately I've been setting gain to 25 and adjusting exposure until the upper end of the histogram sits at about 85-90.

Basic image acquisition process is rough focus, rotate camera to align field to celestial pole, fine focus.  Then I turn on the histogram for SharpCap and pressure tune the scope and try to get the median of the histogram as low as far to the right as possible.  At this point, a majority of the photosphere is being blocked, but it may be necessary to fine-tune it to a certain feature.  I don't think a single-stack is selective enough to make doppler tuning a thing yet.  I'll find out when I can doublestack without having my wife murder me.

After tuning, I adjust the exposure to get it so the upper end of the histogram is around 85-90% of full saturation.  I doublecheck the offset to make sure the black value is as low as I can get it.

 

I grab 1000 images, which corresponds to about 20 seconds of time, probably adequate to avoid motion blur due to an active sun. 
I then stack the images in autostakkert 3.0, and apply the 1.5x drizzle enhancement.  3.0x is too much and there isn't much improvement over 1.5x, but 1.5x does help knock down the grid pattern a bit from the imx178 sensor based cameras while bringing out a bit of subpixel info.

After that, I take the image into imppg and gimp.  I have a video covering what I do in imppg, it is a bit much to describe in text.  I'll link that as soon as youtube finishes processing the video.
 

No photoshop.  I only used Gimp to flip image horizontally and do final level adjustments. This is only a single image, there isn't a prom layer and a disk layer being stacked.  I'm using a trick inside of imppg to pull this off, which is why there is a bright limb.  It is important to note that this is a compromise due to not having a double-stacked scope.  The second etalon will set me back $850 and some change.  My wife is OK with $500... she starts eyeing the larger kitchen knives when I mention $600... and she grabs one of them if I start talking about $750 and up.  lol.gif

I am writing up a start-to-finish guide, mainly for my wife so she can use the scope when I'm not around and get her own images. I'll hopefully be able to share this soon.
 

What I was suggesting was processing your single image twice.  Process it to bring out the proms and save that as a new image.  Then process the original again but just for the surface and save that as a new image.  Then I don’t know if you can do this in Gimp but you would open both images, select the disk of the sun from the surface image and layer that over your prom image and then resize it to match the size of your prom disk.  Maybe that is only possible in PS, I haven’t worked with Gimp before.  The way you have done it works well, I was looking for another trick to help you minimize the bright edge.



#10 MalVeauX

MalVeauX

    Fly Me to the Moon

  • *****
  • Posts: 5294
  • Joined: 25 Feb 2016
  • Loc: Florida

Posted 21 April 2019 - 06:58 AM

Video covering my ImPPG process:

 

https://www.youtube....h?v=0LpiVFcabbo

Good IMPPG tutorial Jason, the part on using Adaptive Unsharp mask was great!

 

Very best,



#11 sushi1128

sushi1128

    Messenger

  • *****
  • Posts: 490
  • Joined: 17 Dec 2013

Posted 21 April 2019 - 07:14 AM

very nice video. i like the simplicity of your technique on using the imppg graph to control surface and prom exposure. tfs



#12 rigel123

rigel123

    Hubble

  • *****
  • Posts: 16356
  • Joined: 29 Jun 2009
  • Loc: SW Ohio

Posted 21 April 2019 - 08:43 AM

Great tutorial!  I'm sure people starting out with imppg will really gain from the techniques you demonstrate.



#13 MAURITS

MAURITS

    Surveyor 1

  • -----
  • Posts: 1947
  • Joined: 22 Sep 2009
  • Loc: Belgium

Posted 21 April 2019 - 01:55 PM

Jason, this is the BEST video I ever saw about Imppg, congrats waytogo.gif

There is so little  information  about this.

This is indeed the technic that I use with Imppg, otherwise it happens that I use Registax RV6 too.


Edited by MAURITS, 21 April 2019 - 02:03 PM.


#14 Great Attractor

Great Attractor

    Mariner 2

  • -----
  • Posts: 250
  • Joined: 23 Oct 2012
  • Loc: Poland

Posted 22 April 2019 - 11:25 AM

That's a great video, Jason, thanks for taking the time to make it! I'd like to add a link to it on the ImPPG's homepage if it's OK.

 

Some thoughts:

 

1) You begin by normalizing to 0-100%. I'd advise against it - any sharpening process will produce output that is slightly below the darkest and above the brightest input values. I.e. you risk losing some data at your darkest and brightest areas.

 

The normalization function was rather meant for batch processing, where you have strong variance in input brightness of your images (i.e. with a Sun time lapse where sometimes there was a high cloud, or like this multi-hour "Moon sunrise" sequence where the sky was getting significantly darker). And then I'd suggest using 10-90% or 20-80% to have a margin for sharpening.

 

For single images, simply doing a "histogram stretch", i.e. moving the leftmost and rightmost nodes of the tone curve so that they encompass the whole histogram, is safer (or just click the "stretch" button). As a reminder, the histogram is calculated *after* sharpening.

 

2) Around 9:53 you say "we folded the histogram over". Were you referring to the fact that the histogram of the current selection now only reaches to about the half of 0-1 range? That's not connected to the tone curve (the histogram shows inputs *before* tone curve application); it's just that this new selection contains the background and not-too-bright disc edge, while the first selection encompassed also the active region with it's white plage areas.

 

3) 15:54 - I'd say the limited detail on the prominence was only due to less-than-optimal seeing, and not due to using a single-stacked scope and a single exposure. The LS50 can pull quite fine details on the proms in good conditions (see e.g. here).

 

4) As for the improved "3D feel" of a negative image I totally agree! Usually filaments have a brighter region just underneath, so after inverting it looks like a proper "shadow". Great especially for large filaments (like this one).

 

5) Since more people are using the "negative disc/positive limb" approach, I guess a tone curve zoom function would be useful for easier manipulation of the dark-end levels. (For now one can just resize the tone curve window). And maybe also selective smooth/sharp curve nodes.

 

One last thing - you can scroll the image by dragging it with the middle mouse button.


Edited by Great Attractor, 22 April 2019 - 11:26 AM.


#15 jwestervelt

jwestervelt

    Viking 1

  • -----
  • topic starter
  • Posts: 554
  • Joined: 01 Sep 2012
  • Loc: Phoenix, Arizona

Posted 22 April 2019 - 01:33 PM

That's fine if you want to use it.  I am still learning the process and I did this mainly to show folks here what I was doing, thus why the video is "unlisted".  I was planning on a second pass after some refinement, and I can certainly take your suggestions into account.

Regarding point #1, I had reservations on this since I didn't see too much documentation, but taking into account what you stated about the histogram stretch and it taking effect after sharpening, I will certainly do that in my second revision.

Regarding point #2, I guess it is a matter of terminology.  The way I look at it, if you have brightness on a scale of 0.0 to 1.0, with 0.0 being black, then the histogram is just a way of viewing the frequency of pixels that contain specific values between 0.0 and 1.0.  What I meant to convey is that the original image would have a range of brightness from 0.0 to 1.0, but that the thin double limb from a single-stacked H-a scope would fall within a very narrow range, say 0.010 to 0.012.  Most of the disk will be back-illuminated by the photosphere and fall above these values, with sunspots being a possible exception.  Likewise outside of the photosphere's disk, few points would be above this range, but it is certainly possible during flaring activity and the process could use some tweaking.  The process that I wanted to describe involves re-mapping the [0.0-0.012] range (prominences and chromosphere limb) to a wider range [0.0-1.0], and then inverting the tone assignment above that range [0.012-1.0] would be remapped to [1.0-0.0].  I don't know what terminology you'd use to refer to this process.  I look at it as "hey, we ramp up the curve, but we start clipping everything inside of the photosphere limb, so we'll just bend/fold the tone curve over so instead of pegging the values at 1.0, we work linearly back to zero".

Regarding point #3, I know that the LS50 can certainly do a better job at grabbing prominences, but the video that you provide as an example clearly is overexposed on the disk.  This video would have the prominences fully contained with the whole range of brightness, whereas in my case where I'm trying to grab both at once, the prominences will only be contained in the first 1-2% of the available 8bit or 16bit space.  Unfortunately we can't exactly "stack" two different exposures when making videos unless we have a very fast and consistent way of jumping between two exposure values with capture software... i haven't seen such a thing in sharpcap or firecap. This is why a double-stacked scope is superior.  I've attempted animations of the proms like this, but transparency has been bad lately and the resulting "haze" around the sun is a bit too much for me. 

Also, regarding timelapse videos, I haven't really found a process for doing this yet.  Manually aligning the resulting images was a huge pain.  I think the process would involve grabbing 5-sec segments of video containing something like 200-300 frames of video, and doing that every 30 seconds.  Meanwhile, you'd have a process that would rank and stack the images in those 5-sec segments to generate a single, stacked image.  Those stacked images would then need to be normalized, aligned, and placed into a video format. It would be amazing if there was a single piece of software that would do all of this.

Thanks for the advice regarding scrolling, i'll try that next time.


 


Edited by jwestervelt, 22 April 2019 - 01:41 PM.


#16 Great Attractor

Great Attractor

    Mariner 2

  • -----
  • Posts: 250
  • Joined: 23 Oct 2012
  • Loc: Poland

Posted 23 April 2019 - 09:49 AM

Great, I will wait for the final version of the video then.

 

As for point #2, got it! (I misunderstood previously).

 

#3 - right, often it's either the disc + barely visible proms; or overexposed disc and prominences in full view. But sometimes, for bright & thick proms based right at the limb, a good single exposure is doable even with single stack, and only slight upping of the tone curve's dark end is needed.

 

Timelapses - hey, that's primarily why I started working on ImPPG, to streamline the whole process. For me it looks like this:

 

- grab a series of videos (sometimes 15 minutes total is enough, the longest I made was ~5 hours), usually 15-30 s long with 15-30 s intervals. I use a homemade capture tool, but working on adding the "series capture" function to Marco's Planetary Imager; also FireCapture can do it.

 

- batch-stack them (can be done in at least AviStack 2 and AutoStakkert!; I use Stackistry) and afterwards check for invalid stacks; for my typical series of 200-300 videos I usually have to correct 10-15 of them, i.e. manually tune stacking settings, change anchor points, etc.

 

- find the best processing settings in ImPPG (try it on a few stacks of the whole series to make sure) and apply them to all the stacks using the ImPPG's batch processing function.

 

- align the resulting processed images using Tools/Align image sequence... in ImPPG.

 

- open as layers in GIMP, if needed crop/rotate/reflect/colorize, preview as animation, export to an animated GIF.

 

Stacking takes most of the time, of course; I hope to push some of its phases onto the GPU eventually.


  • MalVeauX likes this

#17 torsinadoc

torsinadoc

    Apollo

  • *****
  • Posts: 1083
  • Joined: 04 Feb 2013
  • Loc: Fayetteville AR

Posted 23 April 2019 - 03:34 PM

Great video.  The part about adaptive unsharp mask was great!  


Edited by torsinadoc, 23 April 2019 - 03:34 PM.


#18 jwestervelt

jwestervelt

    Viking 1

  • -----
  • topic starter
  • Posts: 554
  • Joined: 01 Sep 2012
  • Loc: Phoenix, Arizona

Posted 23 April 2019 - 05:11 PM

Thanks for the feedback regarding making image sequences, I will try that out.

It is looking like I need to write my own video capture tool.  If I intentionally treat my camera as a color device (it is mono), i can see a separation in the peaks between the red, green, and blue components, meaning that there IS something acting like a bayer mask on this sensor.  Sharpcap and Firecapture lack any ability to manually adjust the sensor that I can tell.  I'll probably need to either find a utility, or make one... and then if the image is captured in color, convert it back to monochrome before processing.  I'm hoping that the manufacturer has a workaround in sofware, but I'm not holding my breath.



#19 jwestervelt

jwestervelt

    Viking 1

  • -----
  • topic starter
  • Posts: 554
  • Joined: 01 Sep 2012
  • Loc: Phoenix, Arizona

Posted 30 April 2019 - 10:23 AM

Still working on this.  I'm fighting the seeing problem at the moment and trying to figure out what is best.  Looking at prom images that Great Attractor posted, I cannot even come close to that with the seeing that we have here in AZ right now.  It is probably just too hot outside.  I might have to try taking a trip to one of the nearby lakes, or maybe go spend a weekend up on one of the mountains.  frown.gif

I tried this the other day, specifically to try an animation of just the proms, and it didn't go so well.  I only got in about 10 minutes of captures before the clouds started to roll in.  The seeing was bad enough that I didn't feel the need to wait the clouds out.

get.jpg?insecure


Edited by jwestervelt, 30 April 2019 - 10:31 AM.

  • Great Attractor likes this

#20 Great Attractor

Great Attractor

    Mariner 2

  • -----
  • Posts: 250
  • Joined: 23 Oct 2012
  • Loc: Poland

Posted 30 April 2019 - 01:44 PM

I might have to try taking a trip to one of the nearby lakes, or maybe go spend a weekend up on one of the mountains.

Phoenix is not that far from the great location I had the pleasure to use (was there on a business trip) - the S18 as it goes up the mountains from San Bernardino, CA and turns east following the ridge at some 6000 ft. Lots of view points, one can break out the imaging setup, enjoy the scenery, pleasant temperature and great daytime seeing (on most sunny days - and most days are sunny). Hey, the Big Bear Solar Observatory (with the currently largest solar telescope - 1.6 m) is in the area for a reason. (I didn't try imaging from Big Bear itself, but I imagine conditions at the lake are also very good.)



#21 MalVeauX

MalVeauX

    Fly Me to the Moon

  • *****
  • Posts: 5294
  • Joined: 25 Feb 2016
  • Loc: Florida

Posted 30 April 2019 - 01:53 PM

Seeing is definitely the limit when it comes to solar-anything. Some places the seeing is so poor most of the time that imaging is very challenging even with such small apertures. Places with really good seeing can make it look like a simple affair, but anyone experiencing poor seeing more often will readily be able to talk about how hard it is to do anything with solar during the day when they want to and the seeing isn't cooperating let alone having a clear sky to do it in.

 

Sometimes animations can be done in a few minutes. Sometimes it takes over an hour. Depends on the structure and how fast it's happening. Some proms don't move quickly. Some are so rapidly moving that they're totally different in less than 10 minutes. Scale matters too, smaller scales will show less change compared to a large scale, granted, the larger scale also demands excellent seeing which is asking a lot.

 

For example, here's an 80 minute animation of a big prom group. It doesn't move much. I was fortunate to have seeing and clarity for that long period of time. This is with 60mm aperture. The surface fibrils and spicules dance around more prominently in the 80 minutes than the big prominence does. Funny how that works.

 

get.jpg?insecure

 

Then, relative to that, here's a prominence animation that is only a few minutes long, maybe 8 minutes or so if I recall, as I was capturing every 30 seconds, showing how rapidly a prominence can appear and change. Granted, this is not a common prominence. But, comparing 8 minutes with one prominence that moves really fast to an 80 minute example where it doesn't hardly move at all, just showing the relative difference, some proms move a lot and some don't. In this one, see how the fibrils and spicules barely move due to how short the time is, yet that prominence is screaming fast.

 

get.jpg?insecure

 

IMPPG did both of these. Thanks again for IMPPG! flowerred.gif

 

Very best,



#22 jwestervelt

jwestervelt

    Viking 1

  • -----
  • topic starter
  • Posts: 554
  • Joined: 01 Sep 2012
  • Loc: Phoenix, Arizona

Posted 30 April 2019 - 04:55 PM

Yeah, imppg is great, but I must have missed an option to align since I had to align manually with my image editor.  I might need to play with it more, or maybe actually RTFM. lol.gif

What I did find interesting about my short animation was on the limb right at 3 o'clock.  You can see 3 proms, and the one in the middle appears to spawn a clone which travels to the one on the bottom.  grin.gif



#23 jwestervelt

jwestervelt

    Viking 1

  • -----
  • topic starter
  • Posts: 554
  • Joined: 01 Sep 2012
  • Loc: Phoenix, Arizona

Posted 30 April 2019 - 05:02 PM

OK, I'm an idiot. :D  I found out how to do it, needs a separate step apparently.  I was thinking batch processing would align as well.  I just tested it and it aligned even better than I did when manually doing it.  bow.gif




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