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

Drizzling discussion

  • Please log in to reply
11 replies to this topic

#1 adamphillips

adamphillips

    Messenger

  • ****-
  • topic starter
  • Posts: 453
  • Joined: 21 Sep 2014

Posted 16 July 2020 - 10:47 AM

Ok so I just drizzled my first image the other day after finally dithering. I don’t know why I put it off so long. It seems to work really well.

From what I can see it’s like a reverse binning. 2x drizzle splits a pixel into 4. 3x drizzle splits 1 into 9.
But I’m thinking it does a better job than just upsizing the number of pixels in photoshop. I would think it’s using the dithered images to find the approximate values for each of the pixels. Am I right?

I processed in drizzled and drizzled and obviously drizzled is better.
Then I wanted to make sure of something. I sampled the processed drizzled image down to the same number of pixels as the nondrizzled and sure enough it was still better.

So that is very cool.
Here’s a weird question. I know many people out there are drizzling but the pixel scale on astrobin is usually the correct one.
Is everybody drizzling and then making their pixels the same size at the end?

If anyone else has anything to add or teach me I want to hear it.
  • elmiko likes this

#2 james7ca

james7ca

    Cosmos

  • *****
  • Posts: 8,630
  • Joined: 21 May 2011
  • Loc: San Diego, CA

Posted 16 July 2020 - 11:29 AM

Drizzling (when done correctly with a dithered set of images) will probably always look better than a simple resampling done after the fact. However, you probably won't see much improvement in detail and sharpness unless you undersampled the original subs. Fortunately, most DSO images are captured at a pixel scale that is probably undersampled, but that may not be true for planetary work where you may already be working close to the diffraction limits of the optics (given relatively good seeing and with lucky imaging techniques).

 

However, drizzling has its drawbacks (it's not a "free lunch") since when you drizzle you are also decreasing your signal to noise (somewhat). Basically, you're trading off some signal to noise to get a higher spatial resolution (similar to what happens when you swap out a camera to one having smaller pixels). So, will a drizzled image with a higher spatial resolution but a slightly decreased signal to noise look better than an undrizzled original? It may or may not, depending upon what you favor. Plus any evaluation done strictly by appearance is likely to be somewhat biased, meaning your eyes may see one thing while a set of measurements may indicate something else. 

 

That said, if you are so far undersampled that your stars appear square or diamond shaped then a drizzle may be your only salvation. (unless you increase your effective focal length or use a camera with smaller pixels or some combination of both).


  • elmiko, t-ara-fan and N5SE like this

#3 elmiko

elmiko

    Aurora

  • *****
  • Posts: 4,836
  • Joined: 27 Jul 2009
  • Loc: Arizona

Posted 16 July 2020 - 03:20 PM

What I have found , drizzle integration makes my stars smoother and less pixelated when you zoom in on the image. It also makes the file size huge!  Funny, when I drizzle my Hyperstar data , which is 2.5 arcsec/ pixel I don't see any difference. But with my Refractor setup it's 1.8 arcsec/ pixel I see quite a difference.



#4 mewmartigan

mewmartigan

    Apollo

  • *****
  • Posts: 1,238
  • Joined: 02 Jul 2008
  • Loc: outside of Pittsburgh, PA

Posted 16 July 2020 - 03:44 PM

I always drizzle with my OSC. As others say, it makes the file huge. At this larger size I will crop/DBE/Photometric color calibration and noise removal. I then resample back to original size for the remainder of processing. 



#5 adamphillips

adamphillips

    Messenger

  • ****-
  • topic starter
  • Posts: 453
  • Joined: 21 Sep 2014

Posted 16 July 2020 - 08:37 PM

I always drizzle with my OSC. As others say, it makes the file huge. At this larger size I will crop/DBE/Photometric color calibration and noise removal. I then resample back to original size for the remainder of processing. 

ive only processed twice with it, I like to process it and get it looking really nice, just about where im happy and then resample it back down.

my logical mind says it will end up looking better doing it that way, but that's just my guess



#6 happylimpet

happylimpet

    Aurora

  • *****
  • Posts: 4,814
  • Joined: 29 Sep 2013
  • Loc: Southampton, UK

Posted 17 July 2020 - 03:19 AM

You dont generally need to dither to be able to drizzle. There'll be enough random movement anyway

 

 

What I have found , drizzle integration makes my stars smoother and less pixelated when you zoom in on the image. It also makes the file size huge!  Funny, when I drizzle my Hyperstar data , which is 2.5 arcsec/ pixel I don't see any difference. But with my Refractor setup it's 1.8 arcsec/ pixel I see quite a difference.

That suggests that the hyperstar system is fundamentally producing poorer resolution than your refractor, which isnt too surprising.


  • elmiko likes this

#7 elmiko

elmiko

    Aurora

  • *****
  • Posts: 4,836
  • Joined: 27 Jul 2009
  • Loc: Arizona

Posted 17 July 2020 - 09:32 AM

I agree, I have successfully drizzled subs that weren't dithered. 



#8 adamphillips

adamphillips

    Messenger

  • ****-
  • topic starter
  • Posts: 453
  • Joined: 21 Sep 2014

Posted 17 July 2020 - 05:30 PM

I agree, I have successfully drizzled subs that weren't dithered. 

ok so heres another million dollar question then,

am I wasting time dithering?

I have an old school ccd I haven't joined the cmos club yet. I actually don't even do darks anymore because I have never seen a difference.

I do bias because it gets rid of a couple hot columns I have, and flats which are huge

 

but I don't have this walking noise people talk about or amp glow.

do I even need to waste the time to dither?

if I don't the image is really not going to move much at all.


  • elmiko likes this

#9 elmiko

elmiko

    Aurora

  • *****
  • Posts: 4,836
  • Joined: 27 Jul 2009
  • Loc: Arizona

Posted 17 July 2020 - 06:17 PM

You will have to make that decision. I only tried drizzle integration once without dithered subs, and it worked. Don't know if I got lucky or not. You could do a test. The next time you are imaging, start your session with say, 20 subs without dithering. Then 20 subs with dithering. Calibrate and Integrate them separately, then evaluate the two stacks side by side.


Edited by elmiko, 17 July 2020 - 06:19 PM.


#10 jdupton

jdupton

    Mercury-Atlas

  • *****
  • Posts: 2,585
  • Joined: 21 Nov 2010
  • Loc: Central Texas, USA

Posted 17 July 2020 - 06:43 PM

Adam (adamphillips),

 

   You can get by without dithering if your polar alignment is off just a little. That imparts a slow drift that will allow drizzling to work. If you note any walking noise, you will probably need to actually dither since that causes a random rather than linear movement in the sub-exposures.

 

   If you do not note a small offset between frames as they come in, your polar alignment may be good enough that the natural drift is not enough. In that case, dithering is more or less required. I have had cases after drift aligning my mount that the sub-exposures only drifted about 10 pixels during a four hour imaging session. Drizzling those frames did not work well at all and hot pixels were a big problem since they never moved enough to be detected as such and thus couldn't be rejected during stacking.

 

 

John


  • elmiko likes this

#11 Peregrinatum

Peregrinatum

    Apollo

  • *****
  • Posts: 1,265
  • Joined: 27 Dec 2018
  • Loc: South Central Valley, Ca

Posted 18 July 2020 - 03:59 PM

I used to drizzle integrate everything, then one project I compared the DI image to the LFC integrated image (this was for over sampled data)... I overlayed one on top of the other and toggled them in PI, and while the DI was "smoother" in appearance there was distinct difference in SNR to my eye... so now I only DI for undersampled data...

 

I may be wrong, and want to be wrong on this issue because of the SGP overhead issues that go along with dithering, but what I have read on here is that dithering helps reduce noise and increase SNR so I dither everything when I image...


Edited by Peregrinatum, 18 July 2020 - 04:00 PM.


#12 spokeshave

spokeshave

    Vanguard

  • *****
  • Posts: 2,302
  • Joined: 08 Apr 2015

Posted 18 July 2020 - 05:46 PM

I agree, I have successfully drizzled subs that weren't dithered. 

Perhaps, but just because drizzle integration produces an image doesn't mean that you are getting the benefit of drizzling. Without dithering, drizzling is really nothing more than upsampling. Drizzling only produces a benefit if the subs are dithered.

 

Tim




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