PixInsight Forum

PixInsight => General => Topic started by: Batch on 2014 June 25 11:54:22

Title: General Question on Drizzle
Post by: Batch on 2014 June 25 11:54:22
Is there any reason why you would not use Drizzle for every Integration or are there specific instances when it should not be used? Thanks.

Title: Re: General Question on Drizzle
Post by: pfile on 2014 June 25 12:20:27
the idea is that if you have undersampled your target, and you have dithered, then you can recover lost spatial information using drizzle.

if you are using a long FL instrument under mediocre skies, then drizzle is probably not useful. however if you happen to be using a camera with large pixels, then it may be useful. it's all about the seeing and the resolution of your system in arc seconds per pixel.

as you might imagine then as the focal length gets shorter it becomes harder and harder to oversample any sky with any camera, and so drizzle starts to become useful.

as for what constitutes under sampling, i think i have only ever seen this mentioned onceā€¦ in stan moore's writings: since photography is a 2D endeavor, the 1D nyquist limit that people often quote for audio is not quite correct for photography. instead of 2x sampling it's more like sqrt(2)*2x to critically sample an image. i am not sure of this figure; juan would know the answer to this i'm sure.

Title: Re: General Question on Drizzle
Post by: NGC7789 on 2014 June 25 13:08:40
To add a concrete example, if you are seeing limited (as most of us are) and the seeing is 3 arc seconds then you would want pixel resolution of 1.1 (3/2.8 in rob's example) to 1.5 (classic nyquist) arcsec per pixel. If you're scope/camera combination delivers a lower pixel resolution (more arcsec per pixel as is mine at 2.9 arcsec per pixel) then you are undersampled (not enough samples). If you have a higher pixel resolution (less arcsec per pixel) then you are oversampled (extra samples).

I hope that adds to clarity instead of taking it away.

Keep in mind that all this is theoretical. If drizzling improves your images enough to justify the extra time then do it. The proof is in the pudding.
Title: Re: General Question on Drizzle
Post by: Juan Conejero on 2014 June 25 14:06:05
If your images are not undersampled, the drizzle algorithm will not improve anything. At the contrary, since a drizzle integrated image always loses some signal (the amount of lost signal is directly proportional to the drizzle subsampling ratio), it will degrade the result in comparison with a regular integration.

An exception to this rule is the Bayer drizzle method (http://pixinsight.com/forum/index.php?topic=7184.0), where drizzle is used as a non-interpolating de-Bayering method with a subsampling ratio of 1.
Title: Re: General Question on Drizzle
Post by: Batch on 2014 June 25 17:19:18
Thanks very much for the answers, its much appreciated.

Title: Re: General Question on Drizzle
Post by: marekc on 2014 June 26 16:39:18
I've got a followup question, concerning how we measure seeing:

What is meant by a phrase like "the seeing is 3 arcseconds"? Am I correct in assuming it means something like the following?...

1) If there were no atmospheric turbulence at all, a star would show up as a diffraction pattern with a central Airy disk surrounded by bright rings that get progressively dimmer outward.

2) Due to seeing, that pattern is almost never what we really get. In real life, we mostly get a "diffraction + seeing" point-spread function that looks roughly like a Gaussian peak. (A Gaussian may not be the ideal approximation of this peak, but it's basically a single mountain-like hump for each star.)

3) If we had a very well-sampled image of one of these star PSFs, we could measure its full width at half maximum. That FWHM is what we mean by the "size of the seeing".

Is that roughly correct? I hear people say things like "I've got two-arcsecond seeing", and I am never quite sure what they mean. The FWHM of a PSF `hump'? The half-flux radius (HFR) of that hump? The tightest double star that could be visually resolved if we replaced the CCD camera with an eyepiece? I'm guessing it's the FWHM of a star image.

- Marek
Title: Re: General Question on Drizzle
Post by: mschuster on 2014 June 26 18:02:44
I think seeing is usually measured as FWHM. Diffraction rings are definitely visible on frames of defocused stars in my tests. It may be that light scatter in the optics and limited resolution and dynamic range tend to hide the rings when in focus. Of course frame measured FWHM usually includes disturbances like tracking errors, temperature dependent focus drift, optical aberrations and scatter, and local seeing (dome). The gurus say getting within 1" FWHM of true seeing (measured on a high pole) means you are doing very well.