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

does this look like a normal DBE for a set of DSLR images

  • Please log in to reply
25 replies to this topic

#1 russellmm

russellmm

    Vostok 1

  • -----
  • topic starter
  • Posts: 172
  • Joined: 02 Aug 2020
  • Loc: Gig Harbor WA

Posted 31 August 2020 - 08:15 PM

Would like some feedback if this looks like what one would expect from a Dynamic Background Extraction in Pixinsight?

 

For one, its clear the bottom right is either hot or overexposed in the images (I am not sure why).

 

going through the editing process gets rid of/ hides most of it, but I am not sure if I should expect something cleaner to begin with

 

I tried shooting this target (M33) both at 15 sec and 30 sec to see if the noise gets better but it seems to be about the same.

 

The attached is a stack of 600 images. Using a Canon 5D Mark IV (unmodded) and 300MM F2.8L lens

 

Regards

Russell

 

 

Attached Thumbnails

  • 2020-08-31 17_58_56-PixInsight.jpg
  • 2020-08-31 18_00_43-PixInsight.jpg


#2 AstroBrett

AstroBrett

    Mariner 2

  • *****
  • Posts: 257
  • Joined: 26 Jan 2009

Posted 31 August 2020 - 08:28 PM

Try using about one fourth as many data points. Having too many samples forces the model surface to conform to the high frequency variations, in other words, the model surface is more crumbled when it should be a relatively smooth surface.

 

Brett



#3 Jim Waters

Jim Waters

    Fly Me to the Moon

  • *****
  • Posts: 5,119
  • Joined: 21 Oct 2007
  • Loc: Phoenix, AZ USA

Posted 31 August 2020 - 08:29 PM

Nope - Have you tried ABE?  What direction is the LP coming from?  What were the DBE settings?



#4 russellmm

russellmm

    Vostok 1

  • -----
  • topic starter
  • Posts: 172
  • Joined: 02 Aug 2020
  • Loc: Gig Harbor WA

Posted 31 August 2020 - 08:41 PM

I have tried using fewer points (I usually use like 10-12 per row) this was just me messing around with more points to see if I could get a different result. So same result with fewer points.

 

Attached is the DBE settings (I left it on one of the marks in the lower right hand corner so you could see the noise sitting on that point)

Also attached is what ABE generates. ABE does not like that hot corner at all...

 

That hot corner does not show up in my other images.

 

One issue is that M33 starts a little lower on the horizon and I am shooting it starting at around 1.3-1.4 Airmasses. I am in a Bortel 5 location but I *am* shooting towards the worst part of my light pollution, which is East. 

Attached Thumbnails

  • 2020-08-31 18_32_52-PixInsight.jpg
  • 2020-08-31 18_34_33-PixInsight.jpg

Edited by russellmm, 31 August 2020 - 08:42 PM.


#5 russellmm

russellmm

    Vostok 1

  • -----
  • topic starter
  • Posts: 172
  • Joined: 02 Aug 2020
  • Loc: Gig Harbor WA

Posted 31 August 2020 - 08:53 PM

This is what i looks like if I use just 31 points

 

 

Attached Thumbnails

  • 2020-08-31 18_50_18-PixInsight.jpg
  • 2020-08-31 18_51_38-PixInsight.jpg


#6 17.5Dob

17.5Dob

    Fly Me to the Moon

  • ****-
  • Posts: 7,084
  • Joined: 21 Mar 2013
  • Loc: Colorado,USA

Posted 31 August 2020 - 09:02 PM

I never go over a tolerance of 0.750...most times I leave it a default.

I normally use just 5-10 points around the perimeter and 2-3 near the center.

But seeing all of your versions, it looks like you had a cloud parked to the right of M33 for quite awhile.

 


  • Paul Garais likes this

#7 russellmm

russellmm

    Vostok 1

  • -----
  • topic starter
  • Posts: 172
  • Joined: 02 Aug 2020
  • Loc: Gig Harbor WA

Posted 31 August 2020 - 09:14 PM

Thanks Dave.

 

I reviewed all the images and scored/weighted them before including them. I did not see any variations in the images that looks like a cloud or mists of a cloud, so I don't know. I'll keep playing around with it and see if I can improve it. 

 

Best Regards

Russell



#8 russellmm

russellmm

    Vostok 1

  • -----
  • topic starter
  • Posts: 172
  • Joined: 02 Aug 2020
  • Loc: Gig Harbor WA

Posted 31 August 2020 - 10:54 PM

interestingly, if I use a radial approach to DBE, I get a much more uniform look, but there still seems to be a lot of Light Pollution glow left in the right portion of the image that makes it unusable 

Attached Thumbnails

  • 2020-08-31 20_51_59-PixInsight.jpg
  • 2020-08-31 20_52_10-PixInsight.jpg

Edited by russellmm, 01 September 2020 - 12:45 AM.


#9 bobzeq25

bobzeq25

    ISS

  • *****
  • Posts: 22,824
  • Joined: 27 Oct 2014

Posted 01 September 2020 - 01:23 AM

What we _really_ need to see is the background image.  That tells you how things are going.

 

I generally use about 10-15 points.  Gradients are large scale, there's nothing to be gained (and much to be lost) by using many points.  It's like setting your PhD2 exposure to 0.5, thinking that will guide better.  In both cases wider sampling intervals (time or space) usefully apply a low pass filter to the data.



#10 russellmm

russellmm

    Vostok 1

  • -----
  • topic starter
  • Posts: 172
  • Joined: 02 Aug 2020
  • Loc: Gig Harbor WA

Posted 01 September 2020 - 02:02 AM

here is a DBE with fewer points selected. I boosted the background to make the Gradient easier to see. If I try to put a point in the lower right hand corner, I would have to put the tolerance way higher to make it not go red because the lower right is so blown out for some reason (the single images don't seem to show any weird gradients unique to the lower right half of the images).

 

Attached Thumbnails

  • 2020-08-31 23_58_42-PixInsight.jpg


#11 russellmm

russellmm

    Vostok 1

  • -----
  • topic starter
  • Posts: 172
  • Joined: 02 Aug 2020
  • Loc: Gig Harbor WA

Posted 01 September 2020 - 02:05 AM

This is about as good as I can get the final image to after post processing it. 

Attached Thumbnails

  • 2020-09-01 00_03_54-PixInsight.jpg

  • dswtan, bobzeq25, 2ghouls and 1 other like this

#12 Paul Garais

Paul Garais

    Viking 1

  • -----
  • Posts: 560
  • Joined: 07 Oct 2018
  • Loc: Sendenhorst, Germany

Posted 01 September 2020 - 02:30 AM

I also recommend to use as few points as possible. I started with tutorials that advised to use x samples per row etc.. That is a bad practice, because it will cause checkerboard structure that is clearly visible in the background. As others said above: There is nothing to gain from more samples.

If you have a linear, straight gradient, you only need two samples.
I look at the image and try to understand where the gradients are before setting the first sample. I start with 2-3 samples and look at the gradient map DBE produces and compare it to what I try to fix. I add single samples if needed and check again.

Since I do so, I did not had a case where more than 10 samples where needed.

—Paul

#13 spereira

spereira

    Aurora

  • *****
  • Moderators
  • Posts: 4,638
  • Joined: 21 Apr 2017
  • Loc: Bedford, NH

Posted 01 September 2020 - 07:10 AM

Moving to DSLR ...

 

smp



#14 RJF-Astro

RJF-Astro

    Viking 1

  • -----
  • Posts: 680
  • Joined: 13 Aug 2018
  • Loc: Zeist, Netherlands

Posted 01 September 2020 - 07:35 AM

This is about as good as I can get the final image to after post processing it. 

Definitely not bad for 4-5 hours. What f-ratio did you use? The thing with STF is that it really stretches your image. And uncooled DSLR images can look bad when stretched hard. So yes, your DBE could use some refinement but I would not overdo it.


  • russellmm likes this

#15 Madratter

Madratter

    Voyager 1

  • *****
  • Posts: 13,050
  • Joined: 14 Jan 2013

Posted 01 September 2020 - 09:03 AM

I go into ABE and DBE in some depth in my unfinished second PixInsight Tutorial. You can see the ABE section here:

 

https://astroimages....ion-part-1.html

 

The DBE section immediately follows. In general, I agree that people tend to use more points when doing DBE than is wise.


  • russellmm likes this

#16 russellmm

russellmm

    Vostok 1

  • -----
  • topic starter
  • Posts: 172
  • Joined: 02 Aug 2020
  • Loc: Gig Harbor WA

Posted 01 September 2020 - 09:51 AM

I am using F2.8. 

 

I have tried 15 sec exposures to try and bring down the glow/noise

 

Thanks for the advice all. 

 

I'l take a look at your Tutorial Madratter.


Edited by russellmm, 01 September 2020 - 09:53 AM.


#17 fmeschia

fmeschia

    Apollo

  • *****
  • Posts: 1,482
  • Joined: 20 May 2016
  • Loc: Mountain View, CA

Posted 01 September 2020 - 10:36 AM

Would like some feedback if this looks like what one would expect from a Dynamic Background Extraction in Pixinsight?

 

For one, its clear the bottom right is either hot or overexposed in the images (I am not sure why).

 

going through the editing process gets rid of/ hides most of it, but I am not sure if I should expect something cleaner to begin with

 

I tried shooting this target (M33) both at 15 sec and 30 sec to see if the noise gets better but it seems to be about the same.

 

The attached is a stack of 600 images. Using a Canon 5D Mark IV (unmodded) and 300MM F2.8L lens

 

Regards

Russell

It seems to me that you sampled points in the galactic halo of M33, and as a result DBE overcorrected the gradient, creating a “hole” in which M33 now sits.

 

There’s a great free video by Adam Block that faces a similar image situation, and solves it very nicely with DBE: https://www.youtube....h?v=f6IN9cXVh-4


  • russellmm likes this

#18 bobzeq25

bobzeq25

    ISS

  • *****
  • Posts: 22,824
  • Joined: 27 Oct 2014

Posted 01 September 2020 - 11:33 AM

here is a DBE with fewer points selected. I boosted the background to make the Gradient easier to see. If I try to put a point in the lower right hand corner, I would have to put the tolerance way higher to make it not go red because the lower right is so blown out for some reason (the single images don't seem to show any weird gradients unique to the lower right half of the images).

That background looks correct, a nice smooth gradient.

 

And the final image is nice.


  • russellmm likes this

#19 Euripides

Euripides

    Ranger 4

  • -----
  • Posts: 304
  • Joined: 15 Aug 2019
  • Loc: Thessaloniki - Greece

Posted 02 September 2020 - 09:01 AM

Guys I always look first the background result of DBE or ABE and I can see the patterns. Usually I take max 20 samples and tried different settings, for example Tolerance 6, Sample Radius 100, Minimum Sample Weight 0.150.

 

But I can clearly see "black patches" at my final stages of post. Is this something that can be eliminated with better DBE, or it is just limitation of my gear (stock DSLR, high temps etc.)

 

 

This crop is just before my final DeNoise .

 

 

PS. Those patches are visible even in 12 hours total subs.

 

blackpatches.jpg


Edited by Euripides, 02 September 2020 - 09:04 AM.


#20 Paul Garais

Paul Garais

    Viking 1

  • -----
  • Posts: 560
  • Joined: 07 Oct 2018
  • Loc: Sendenhorst, Germany

Posted 02 September 2020 - 10:24 AM

Guys I always look first the background result of DBE or ABE and I can see the patterns. Usually I take max 20 samples and tried different settings, for example Tolerance 6, Sample Radius 100, Minimum Sample Weight 0.150.

But I can clearly see "black patches" at my final stages of post. Is this something that can be eliminated with better DBE, or it is just limitation of my gear (stock DSLR, high temps etc.)


This crop is just before my final DeNoise .


PS. Those patches are visible even in 12 hours total subs.

attachicon.gifblackpatches.jpg

The tolerance and sample radius seem to be extremely high. I never used values that big, so can not tell if they cause issues.

The patches you are talking about are normal and likely the result of too strong denoising the background and stretching too far. But hard to tell without knowing your exact workflow.
How does the image look if you only stretch it slightly after DBE and without applying denoise or anything else?

#21 RJF-Astro

RJF-Astro

    Viking 1

  • -----
  • Posts: 680
  • Joined: 13 Aug 2018
  • Loc: Zeist, Netherlands

Posted 02 September 2020 - 11:02 AM

Could be a combination of noise reduction and not enough dithering. So it starts with fixed pattern noise and then turns into patches with noise reduction.

#22 Euripides

Euripides

    Ranger 4

  • -----
  • Posts: 304
  • Joined: 15 Aug 2019
  • Loc: Thessaloniki - Greece

Posted 02 September 2020 - 04:24 PM

This is a comparison between DBE and ABE , both with the same small stretch. I believe that DBE is better.

 

 

DBEvsABE.jpg

 

 

 

Here are shots with my workflow

 

https://www.dropbox....MuMKtzA6fa?dl=0

 

I believe that star reduction amplifies the issue.

 

 

PS. I've included the picture above into the folder.

 

 

 


  • Paul Garais likes this

#23 Paul Garais

Paul Garais

    Viking 1

  • -----
  • Posts: 560
  • Joined: 07 Oct 2018
  • Loc: Sendenhorst, Germany

Posted 03 September 2020 - 07:28 AM

This is a comparison between DBE and ABE , both with the same small stretch. I believe that DBE is better.

 

 

attachicon.gifDBEvsABE.jpg

 

 

 

Here are shots with my workflow

 

https://www.dropbox....MuMKtzA6fa?dl=0

 

I believe that star reduction amplifies the issue.

 

 

PS. I've included the picture above into the folder.

Your DBE is fine, in my opinion. As RJF-Astro says: Your main issue is the noise. In your images prior to noise reduction you can clearly see some diagonal structures in the noise pattern. You can avoid this by dithering. Or, as I learned here on cloudynights: Dither or die!

 

Your noise reduction is definitly the cause for the issues/patches you get. The best advice regarding noise reduction, I can give: Do not kill the noise and every single ancestor and descendant of it. It is called "reduction" not "wipe out" ;). Noise is not bad, it helps to keep a natural look to an image. But it needs to be reduced to a point where it is not dominant anymore. So the best approach is, to generate as less noise as possible while shooting an image. Dithering is one part of avoiding noise.



#24 Euripides

Euripides

    Ranger 4

  • -----
  • Posts: 304
  • Joined: 15 Aug 2019
  • Loc: Thessaloniki - Greece

Posted 03 September 2020 - 07:55 AM

Ok, so then I’m dying cause I do not dither :-)

Probably I have to check again EKOS and give it a second chance.


Sent from my iPhone using Tapatalk
  • Paul Garais likes this

#25 Euripides

Euripides

    Ranger 4

  • -----
  • Posts: 304
  • Joined: 15 Aug 2019
  • Loc: Thessaloniki - Greece

Posted 20 September 2020 - 10:01 AM

Ok so...I took 10 hours of M31 with PHD2 & dither.

 

306 x 2' @f3.2 , ISO100 with SkyTech LPRO Max filter from bortle 8-9, with flats & bias. Darks made the final result worse, so DEL.

 

But probably I am doing something wrong...

 

I've tried multiple times DBE and ABE but no matter what, I always get those ugly patches. I've also minimised my denoise.

 

If anyone had the time to take a look, this is the final stack cropped.

https://www.dropbox....QIZ9PukYNa?dl=0

 

 

I am getting better with my actual targets but I have certainly issues with my background :-( Any tips for the DBE points would be more than welcome!

 

 

M31.jpg


Edited by Euripides, 20 September 2020 - 10:02 AM.



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