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

Pixinsight Drizzle Algorithm Yes/No?

  • Please log in to reply
13 replies to this topic

#1 GeneralT001

GeneralT001

    Vanguard

  • *****
  • topic starter
  • Posts: 2438
  • Joined: 06 Feb 2012
  • Loc: Nanaimo, BC

Posted 01 August 2016 - 12:43 PM

I was wondering if you should always be applying the Pixinsight Drizzle Algorithm during stacking or only sometimes. I understand it doubles the resolution - but is it needed?

 

Are there any guidelines for using this algorithm?

 

 

 

Thanks



#2 torsinadoc

torsinadoc

    Apollo

  • *****
  • Posts: 1088
  • Joined: 04 Feb 2013
  • Loc: Fayetteville AR

Posted 01 August 2016 - 12:59 PM

I made the mistake to drizzle a 6 panel mosaic. The file was massively large.

#3 Jon Rista

Jon Rista

    ISS

  • *****
  • Posts: 23593
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 01 August 2016 - 01:04 PM

Drizzling will only be useful when your data is undersampled. If you have oversampled data, then you are not likely to notice much in the way of resolution gains by drizzling. However undersampled data can benefit, and generally the more undersampled you are, the more likely you are to 



#4 GeneralT001

GeneralT001

    Vanguard

  • *****
  • topic starter
  • Posts: 2438
  • Joined: 06 Feb 2012
  • Loc: Nanaimo, BC

Posted 01 August 2016 - 01:09 PM

Drizzling will only be useful when your data is undersampled. If you have oversampled data, then you are not likely to notice much in the way of resolution gains by drizzling. However undersampled data can benefit, and generally the more undersampled you are, the more likely you are to 

Hey Jon,

 

That makes sense.

 

 

Thanks



#5 spokeshave

spokeshave

    Vanguard

  • *****
  • Posts: 2033
  • Joined: 08 Apr 2015

Posted 01 August 2016 - 01:19 PM

Drizzling will only be useful when your data is undersampled. If you have oversampled data, then you are not likely to notice much in the way of resolution gains by drizzling. However undersampled data can benefit, and generally the more undersampled you are, the more likely you are to 

I would also add that drizzling only works on data sets that are dithered, and only works well on rather large sets of subs.

 

TIm


  • xiando likes this

#6 Jon Rista

Jon Rista

    ISS

  • *****
  • Posts: 23593
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 01 August 2016 - 03:42 PM

Sorry guys...IÁm having internet problems today. There was a lot more to my post...but it seems to have been lost somewhere along the line. I did mention dithering before, as it is critical. And for drizzling to be effective, you really need to make sure you drizzle well enough that the stars are over different pixels in every sub. So, while I sometimes advocate for dithering every few subs...if you intend to drizzle, you want to dither every sub. (Important point that was lost in my original postᴔ)



#7 Hawkdl2

Hawkdl2

    Viking 1

  • *****
  • Posts: 674
  • Joined: 22 Feb 2015
  • Loc: Claremont, CA

Posted 01 August 2016 - 08:52 PM

Here is a comparison of a very zoomed-in section of the un-drizzled and drizzle output of  442 ASI1600 Ha frames dithered through SGP/PHD2 and integrated in PI.  Drizzling was 2x in with a 0.9 drop shrink.  Noise improved about 15% in the drizzled image from 1.177e-5 to 1.020e-5, but visually the drizzled image looks to have a much higher resolution to me. I always seem to notice that drizzling has a big visual impact on stars and structures but the background always seems to look unchanged or maybe even less smooth.

 

(A tweak to my RC6's collimation and my stars are now much rounder than in this example).

 

 

Attached Thumbnails

  • 2016-08-01.jpg

  • AstroGabe and bobzeq25 like this

#8 Jon Rista

Jon Rista

    ISS

  • *****
  • Posts: 23593
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 01 August 2016 - 09:29 PM

Larry, measure the FWHMs in both images. That will actually give you a read on your actual data resolution. While the total pixel count representing each star has increased, my guess is there is not going to be much improvement in FWHM. You were already well sampled in the original image...after drizzling, your radically oversampled. 



#9 cuivienor

cuivienor

    Surveyor 1

  • -----
  • Posts: 1922
  • Joined: 06 Oct 2010
  • Loc: Tokyo, Japan

Posted 02 August 2016 - 05:01 AM

On undersampled data (like what I took with binning 4x4 on my 1600MM), drizzle is magic. 'Nuff said :-)

 

Which leads me to a question: would it be OK to be undersampling systematically to get more shorter subs with less wind effect on stars? How much do I lose by undersampling then drizzling?

Attached Thumbnails

  • Drizzle_BeforeAfter.png

Edited by cuivienor, 02 August 2016 - 05:08 AM.

  • Charlie B likes this

#10 WesC

WesC

    Fly Me to the Moon

  • *****
  • Posts: 5320
  • Joined: 06 Feb 2013
  • Loc: La Crescenta, CA

Posted 02 August 2016 - 10:23 AM

I've been using it with my FSQ-106 data @3.46" (when I get enough subs)... Seems to improve image quality but at the expense of much more resource use. If you've got the disk space, ram and horsepower to do it, and you're under sampled, it can make a positive difference.

 

But if you're not seeing square stars at 1:1... I wouldn't bother. 



#11 Charlie B

Charlie B

    Apollo

  • -----
  • Posts: 1442
  • Joined: 22 Mar 2008
  • Loc: Sterling, Virginia

Posted 02 August 2016 - 04:18 PM

On undersampled data (like what I took with binning 4x4 on my 1600MM), drizzle is magic. 'Nuff said :-)

 

Which leads me to a question: would it be OK to be undersampling systematically to get more shorter subs with less wind effect on stars? How much do I lose by undersampling then drizzling?

Beautiful example of drizzle power, showing rectangular pixels being resolved to round stars.  Deliberately undersampling to use drizzle may not lose anything but your processing will surely expand.

 

Regards,

 

Charlie B



#12 xiando

xiando

    Fly Me to the Moon

  • *****
  • Posts: 6386
  • Joined: 27 May 2015
  • Loc: Cloudy NEOhio

Posted 02 August 2016 - 04:30 PM

I'd note that dithering need not be intentional. As long as the data set approximates a Gaussian positional deviation (a tacit requirement for valid drizzle) with it can still be used effectively.



#13 rflinn68

rflinn68

    Fly Me to the Moon

  • *****
  • Posts: 6080
  • Joined: 09 Mar 2012
  • Loc: Arkansas

Posted 02 August 2016 - 08:32 PM

I recently posted my results, in another thread, on drizzling my 2X2 SII data in the Sadr region. The data was 40X900s at 3.16"/pixel (2X2). Drizzle works magic with stuff like this. I just use a small dither between subs.

 

Drizzle_Comparison.JPG

 

I also just drizzled my 1.12"/pixel data with the 10" RC (2X2). It was the initial testing of my carbon truss RC and it was all shot at 2X2. My primary reason for dithering that was to get a larger final image. It didnt help much, if any, but didnt hurt anything either, and it got me a larger image for display purposes. Interesting experiment anyway.



#14 cuivienor

cuivienor

    Surveyor 1

  • -----
  • Posts: 1922
  • Joined: 06 Oct 2010
  • Loc: Tokyo, Japan

Posted 02 August 2016 - 10:32 PM

Here is what drizzle gives me on the same set of narrowband data, after combination.

Attached Thumbnails

  • post-147653-0-09257700-1469853183.jpg
  • M16Narrowband_Green.jpg



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