Author Topic: Salvage an old integration of Arcturus (remove reflection)  (Read 7040 times)

Offline steven_usa

  • Member
  • *
  • Posts: 74
    • View Profile
I'm looking for some help to remove a bad "gradient" that is being caused by some kind of reflection from the secondary mirror of the scope used (VMC200L).  The sensor was from a Canon 550/T1i DSLR.  The image itself isn't so important to me, but I was curious about techniques to help remove these kinds of effects (I assume somehow caused by the focuser not being perfectly centered?).

The integration consisted of 19 out of 30 collected lights at 60", ISO3200 (probably could have used all 30, but I was exercising some scoring selection approaches outlined by David Ault -- that got me down to 26 Lights, and then only 19 Registered).  A superbias from ~30 bias images was used.  The darks were taken at the incorrect duration (10sec instead of 60sec), but I used them anyway in the ImageCalibration.  And the lights were taken at an incorrect exposure of 1/2000 (so they have SOME data, but barely).

No dithering.

This data was collected over a year ago, long before I had PI.  What's interesting to me is that DSS makes a reasonable result from this data, but PI reveals the reflection and I simply can't scrub it away (e.g. with ADE or BDE) and get any reasonable result.   

Here is the integration file (86mb, 16-bit XISF):  https://app.box.com/s/12uhy7dehsafuskv9g8iim893lreur4f   (should be ready to stretch: had Canon Band Reduction applied,  Repaired HSV Separation,  Background Neutralization and a few other initial steps -- this is a PI specific integration and does NOT reflect the same set of calibration files as used in the DSS example result depicted below)

For comparison (BELOW), attached below is my DeepSkyStacker result with minor adjustments in CS6 (in some monitors, keen observers will still make out the faint reflection even in this result).    Also note that AstroBin doesn't plate solve the image.




« Last Edit: 2015 October 18 11:28:15 by steven_usa »

Offline steven_usa

  • Member
  • *
  • Posts: 74
    • View Profile
Re: Salvage an old integration of Arcturus (remove reflection)
« Reply #1 on: 2015 October 17 07:11:07 »
And for reference here is an example of the integration stretched in PI, as a preview of the reflection I'm talking about  (curious if anyone else has any suggestions on the cause of the reflection and/or suggestions for cleaning it up -- though I admit ISO3200 and lack of dithering might not really make it worth it on such an easy target, but it's an exercise/challenge :) )


Offline oldwexi

  • PixInsight Guru
  • ****
  • Posts: 622
    • View Profile
    • Astronomy Pages G.W.
Re: Salvage an old integration of Arcturus (remove reflection)
« Reply #2 on: 2015 October 17 10:47:21 »
If you run a STF over your
DSS CS Autosave002_adjusted.jpg
Output, it shows the same reflections.

So, such a bright star gets you more  reflections the more glass is in the Image Train.
No matter what software you use...

Would use PixelMath for creating a ring mask over the reflection ring and try to carefully reduce the brightness of the reflection.
create a ring with the PixelMath expression:          inellipse(430,382,311,311) - inellipse(430,382,145,145)

Gerald

Offline steven_usa

  • Member
  • *
  • Posts: 74
    • View Profile
Re: Salvage an old integration of Arcturus (remove reflection)
« Reply #3 on: 2015 October 17 14:07:34 »
I've noticed NOT including the Darks in the stack seems to improve things.  However, then I get a problem with StarAligment and may have to use tedious  DynamicAligmment instead.  I'll try to complete a sans-Dark integration...
« Last Edit: 2015 October 17 15:27:51 by steven_usa »

Offline oldwexi

  • PixInsight Guru
  • ****
  • Posts: 622
    • View Profile
    • Astronomy Pages G.W.
Re: Salvage an old integration of Arcturus (remove reflection)
« Reply #4 on: 2015 October 17 14:25:51 »
What is improved not using Darks???

Gerald

Offline steven_usa

  • Member
  • *
  • Posts: 74
    • View Profile
Re: Salvage an old integration of Arcturus (remove reflection)
« Reply #5 on: 2015 October 17 16:21:43 »
I was mistaken.  A single integrated frame looked "better" without the Dark (i.e. less apparent reflection), but the more completed registration+integration I guess still amplifies the effect.   Still (to try without Darks for comparison), of the 30, I've isolated down to 8 and used DynamicAligment on those.

The alternate integration without the Darks is here (VER2):  https://app.box.com/s/rwtkmna24mqth659gnqeutjpucgr2dvf
(8x60", ISO3200, but same Light+SuperBias as before)   It's slightly more cropped, but I widdled an ellipse mask down using:
inellipse(2590,2228,1745,1745) - inellipse(2590,2228,775,775)

Though I'm frankly not sure how to dial down the brightness (or certainly without reducing a portion of the star also).

Below is the STF Stretch with this alternate integration (plus a few steps I made after the initial integration, like some ABE/SCNR) and an example of what the mask looks like (not inverted since I found that easier to help guide centering the ellipses).


NOTE: Data was taken 2014-06-06 (Texas summer) and the DSLR sensor was sizzling at 38degC even at 1:30am.  Arcturus was setting at about 50deg above the horizon to the West (with the moon just barely crossing the horizon almost directly due West, probably even below the fence line).    No LP filters were in the imaging train (the T1i/550D camera was stock LPF1 and 2).
« Last Edit: 2015 October 17 22:20:33 by steven_usa »

Offline slang

  • Member
  • *
  • Posts: 56
    • View Profile
Re: Salvage an old integration of Arcturus (remove reflection)
« Reply #6 on: 2015 October 17 22:52:52 »
Hiya.

If you run a STF over your
DSS CS Autosave002_adjusted.jpg
Output, it shows the same reflections.

So, such a bright star gets you more  reflections the more glass is in the Image Train.
No matter what software you use...

Would use PixelMath for creating a ring mask over the reflection ring and try to carefully reduce the brightness of the reflection.
create a ring with the PixelMath expression:          inellipse(430,382,311,311) - inellipse(430,382,145,145)

Gerald

Wow - that's a brilliant idea. I've got one image from a long time ago that I am keen to reduce a reflection on. Are you aware of a way to 'feather' the edge so they're not so high contrast?
--
Mount: Orion Atlas 10 eq-g, GSO RC8, Astrophysics CCDT67
Guiding: ST80/QHY5L-II-M
Cam: Canon EOS 450D (IR Mod), QHY8L

Offline slang

  • Member
  • *
  • Posts: 56
    • View Profile
Re: Salvage an old integration of Arcturus (remove reflection)
« Reply #7 on: 2015 October 17 22:54:58 »
Hey.

I was mistaken.  A single integrated frame looked "better" without the Dark (i.e. less apparent reflection), but the more completed registration+integration I guess still amplifies the effect.   Still (to try without Darks for comparison), of the 30, I've isolated down to 8 and used DynamicAligment on those.

My guess would be that with the dark noise left in place (i.e. no dark subtraction) then you'd have more even noise across the frame, which would make all features less noticeable, including the reflection.

Cheers -
--
Mount: Orion Atlas 10 eq-g, GSO RC8, Astrophysics CCDT67
Guiding: ST80/QHY5L-II-M
Cam: Canon EOS 450D (IR Mod), QHY8L

Offline msmythers

  • PTeam Member
  • PixInsight Jedi
  • *****
  • Posts: 1176
    • View Profile
    • astrobin
Re: Salvage an old integration of Arcturus (remove reflection)
« Reply #8 on: 2015 October 17 23:52:50 »
You can use the Convolution tool to soften the edges.


Mike

Offline slang

  • Member
  • *
  • Posts: 56
    • View Profile
Re: Salvage an old integration of Arcturus (remove reflection)
« Reply #9 on: 2015 October 18 00:41:42 »
You can use the Convolution tool to soften the edges.

Mike

Fantastic, thanks. I was having trouble with this (wasn't working on my 'mask') as it was an RGB image. Extracted L, and convolution worked as you show here.

Thanks again.

Cheers -
--
Mount: Orion Atlas 10 eq-g, GSO RC8, Astrophysics CCDT67
Guiding: ST80/QHY5L-II-M
Cam: Canon EOS 450D (IR Mod), QHY8L

Offline steven_usa

  • Member
  • *
  • Posts: 74
    • View Profile
Re: Salvage an old integration of Arcturus (remove reflection)
« Reply #10 on: 2015 October 18 08:29:53 »
As a reminder, the "Darks" that I had were only 10s, so they didn't match the 60" duration of the Lights (and those Darks were about 2degC cooler, being 36-37degC as opposed to 39-40degC -- temps as reported by the non-cooled DSLR internal sensor).     (if I still had the camera I might just try to take new Darks)

But agreed, the original version with those Darks is adequate or certainly doesn't look worse.  Here they are side by side with 24-bit and Boosted STF to emphasize the resulting noise  (I was experimenting with a little bit of Decon on the NoDarks version at the time, it's not in the actual initial integration)


Wish we could integrate a touch-tablet and literally fat-finger where to apply (or not-apply) the mask :)

Offline oldwexi

  • PixInsight Guru
  • ****
  • Posts: 622
    • View Profile
    • Astronomy Pages G.W.
Re: Salvage an old integration of Arcturus (remove reflection)
« Reply #11 on: 2015 October 18 09:38:21 »
Hi Steven!
These reflections are LargeWavelets artefacts.
They are better to remove with MultiscaleMedianTransform.

But, before i used this process i have softened the PixelMath generated mask with the process RangeMask by setting the smoothness
to 10.0 !  Thats for me more easy in this B/W mask case  than with Convolution as Range_Mask softening has only one slider.
Have also changed a little the radii of the ring in PixelMath.

So, using the softened mask you run the process Multiscale MedianTransform.
Set the Layers to 8, set the Bias of the R (Residual 256)   to  -0.003 execute the process
and the reflection is mostly gone without interaction on the star, because the star is a SmallWavelets target and
therefore stays untouched!

Find the screenshot here: http://www.werbeagentur.org/oldwexi/Reduce_Reflections.jpg


Gerald
« Last Edit: 2015 October 18 09:43:46 by oldwexi »

Offline steven_usa

  • Member
  • *
  • Posts: 74
    • View Profile
Re: Salvage an old integration of Arcturus (remove reflection)
« Reply #12 on: 2015 October 18 11:27:26 »
Very nice!

I guess I'm confused about PixelMath.  If you used the 430x382 center, how did you scale it to the integrated image (and offset it, such that the bottom edge is clipped)?


I tried these steps:
1- Duplicate(clone) the original initial_integration XISF
2- Highlight the duplicate image, select PixelPath process, use inellipse (I'm using my approx. 2500x2228 centered version), Apply (to the cloned image)
3- apply Convolution (or the RangeSelection) as suggested
4- highlight the original initial_integration image, and select the _clone as the mask (or the range_mask) inverted
5- use the MultiscaleMedianTransform as recommended

And I'm not getting your same result, so I'm missing something :)

Thanks!!

Offline oldwexi

  • PixInsight Guru
  • ****
  • Posts: 622
    • View Profile
    • Astronomy Pages G.W.
Re: Salvage an old integration of Arcturus (remove reflection)
« Reply #13 on: 2015 October 18 11:53:45 »
Steven -
Why did you invert the mask?
It has to be a white ring. Inverting the white ring to a black ring
supresses the MMT.

The ring in the small image is not scaled and clipped!!!  If you enter the correct
radius and x,y positions it is  positioned correctly and clipped automatically...

I measured with the ReadoutCursor the width and starting positions of the inner and outer circle.
The width of the outer circle  divided by 2 is the radius.
The horizontal startingpoint of the circle plus radius gives the x coordinate. etc....
(This calculations despite my Age of 69 today  i could do without computer)
Same process for the Y coordinate.

The difficulty to get the same results maybe is - i have only an 8-bit image -
you use, so i hope  a 32-bit image which may be needs adjusting of the Parameter i used.

Do you have the 32-bit Fits available (final  with dark, -flat and -bias calibrated and integrated)
so i can try a better example and advice for using the right parms.

Gerald

Offline steven_usa

  • Member
  • *
  • Posts: 74
    • View Profile
Re: Salvage an old integration of Arcturus (remove reflection)
« Reply #14 on: 2015 October 18 17:23:57 »
The XISF's in the links above should have been saved as 16-bit, was it not?

I've re-ran ImageIntegration again with 15 out of the original 30 Lights.   I resaved that as 32-bit and uploaded here (VER3):  https://app.box.com/s/kxwrb99y8pus7s9yk46zfweooeqcrpwu   (this remains similar to VER1 with the superbias, flats, darks)
note: the reflection noise looks less apartment to me in this 3rd version (but I've done no crop or any adjustment, leaving it exactly as ImageIntegration produced it)

I'm trying to upload all the Lights in registered form, but each one is 177MB and taking awhile.  That folder is here:
https://app.box.com/s/umf0wi6ulsxizfxy965lrhntdkg1zin8  (the integrated 32-bit save is there, but the individual debayered/registered lights will take awhile)


This whole working folder is at:  https://app.box.com/s/8q5ptrau3i3xy16hdwslopw5rpqg5cg9


EDIT: I think my superbias was saved in 16-bit, then applied to 32-bit integration files. Could this be a problem?  (I generally always saved in 32-bit, but was advised for the Canon DSLR image files there really was no need -- so I was going to try processing exclusively in 16-bit to see if I noticed any differences {for all my data, not just this sample in particular})

To recap on the variations I have:
VER1:   initial_integration_Arcturus19x60_ISO3200.xisf    https://app.box.com/s/12uhy7dehsafuskv9g8iim893lreur4f  (should be a 16-bit)
VER2:   8x60_ISO3200_Arcturus_initial_NoDarks_integration.xisf    https://app.box.com/s/991iffw4mxo8iyw2sf14p9xqezxcap96  (16-bit, should be attempt to see how removing Darks makes a difference)
VER3:   VER3_Arcturus_baseline_integration_15x60_ISO3200.xisf  https://app.box.com/s/kxwrb99y8pus7s9yk46zfweooeqcrpwu  (redo of VER1 saved as 32-bit and with a tighter selection of Lights)
« Last Edit: 2015 October 18 18:08:52 by steven_usa »