Minimum Integration Time for Drizzle? - Forums

Topic starter

Nov 27, 2020

I have noticed that when I drizzle vs not doing so(yes, I dither), my drizzled images lack definition and do not look nearly as crisp/good as those that that didn’t drizzle.

Is this due to low integration time (e.g. 3-7 hours)? If so, what is your minimum integration time to achieve desired results with drizzle?

Thanks!

Wei-Hao Wang

Nov 28, 2020

I don't think there is a minimum integration time for drizzle to work.

However, there is a minimum number of dithered exposures for drizzle to work.

If you use a mono sensor and if you want to apply 2x2 drizzle, the absolute minimum is four exposures, if you can accurately control your dither offsets to be +/- 0.5 pixels. In reality, such kind of dither control is not possible (at least for amateur telescopes). So it requires many more exposures. My personal minimum is 10. Many people say you will need more than 30. I think they are exaggerating, but it is true that drizzle will work much better if you have 30 exposures rather than just a dozen.

Well Written Helpful Insightful Concise

schmaks

Topic starter

Nov 28, 2020

Wei-Hao Wang:
but it is true that drizzle will work much better if you have 30 exposures rather than just a dozen

Interesting. For instance, my last session was 95 exposures and yet still the details
in the non-drizzled version are far weaker with the same settings otherwise.

perhaps this is to do with my dslr vs a astro cam?

Wei-Hao Wang

Nov 28, 2020

That means your problem is not insufficient integration nor insufficient number of exposures.

Drizzle works on DSLRs. For example, this image of mine combines Bayer drizzle (called differently in different programs) and traditional 2x drizzle, and has very high resolution. However, for DSLR drizzle to work, the program needs to do it correctly, i.e., has to take the GRGB Bayer pattern into account. Otherwise it will not produce meaningful effects.

In general, for drizzle to work, there are many things that you (or the processing program) have to do right. I think to this point, you have to show us the actual images for us to judge what may have gone wrong.

Helpful Insightful Respectful

dkamen

Nov 28, 2020

Also, "less crisp" is very subjective. Generally speaking, you need to resize your integration at 2X with traditional means (e.g. Photoshop) and compare that with the drizzled version. Otherwise it's an unfair comparison, the 1X image always has a tremendous advantage in contrast and SNR.

Comparison is best done at 100% magnification or more, so you can see tiny details as this is the scale drizzle works at.

If a tiny star is square in the normal integration but circle or oval in the drizzled, this is actually what drizzle does. But subjectively the latter might look more blurry to the eye.

Also, drizzle *always* results in more noise. Depending on the subject and acquisition details it could very well be that the increase in resolution does not compensate for the increase in noise, IOW the drizzled image is worse.

Please bear in mind that drizzle is no magic wand. You will never do better than the theoretical optimum allowed by your optics and seeing. It only allows you to go closer to it if your pixel size is not up to the task. For example, if you optics on a clear night can resolve 0.5 arcsec but your pixel is 2 arcsec wide, drizzle might give a passable result with 1 arcsec resolution. But if your seeing and optics are 2.5 arcsec (not unlikely at all) drizzle cannot improve anything as you are already very well sampled. On the contrary, it will mostly add noise and reduce contrast.

Cheers,
Dimitris

Helpful Insightful Engaging

dkamen

Nov 28, 2020

* bare in mind (😳)

Bogdan Borz

Nov 28, 2020

Hi schmaks,

Drizzle does not depend on time. But the question is do you need to drizzle? Drizzle is required when you want to increase the resolution, if your stars fall on too few pixels (hence they look square). So you can inspect your frame and zoom in to see the small stars shapes. Secondly, you can use a function in Pixinsight called FWHMEccentricity script. It will measure the descriptive statistics for your stars. If your Median FWHM is < 2 pixels (meaning 50% of your stars have a FWHM less than 2 pixels), you are going to benefit from Drizzle.

Secondly, remember that if you will work on the Drizzle master, you have to increase your scale for all the processes (I am talking about Pixinsight), since the pixel number has doubled (if you sharpen at 3 pixels radius you're not going to see the same effects as on the initial image. you have to double it).

CS !

Helpful Engaging Supportive