Darker image after DrizzleIntegration

[kentnek]

Having used mainly DSS for stacking, I'm currently learning how to stack with PI using the WBPP script.

I'm also experimenting with DrizzleIntegration, which produces a image a tad cleaner than the original master light, but slightly dimmer and less contrast-y (especially around the fainter Horsehead Nebula region):

left is master light, right is drizzled, only autostretch + ABE applied

With my highly undersampled setup (4.51"/pixel), I think drizzling would be beneficial. However, it seems like the SNR is lower after doing so. Is this expected, and which image should I continue my editing process on? Any help/suggestion is highly appreciated!

Edited January 10, 2022 by kentnek

[vlaiv]

3 minutes ago, kentnek said:

With my highly undersampled setup (4.51"/pixel), I think drizzling would be beneficial. However, it seems like the SNR is lower after doing so.

I'm highly skeptical that you are highly undersampled

In principle, if you are really undersampled - you can trade some SNR for sharpness by using drizzle, but I think that it never really works in amateur setups. Drizzle was developed for Hubble - which is in outer space and can be pointed precisely. It is used to circumvent undersamplig of telescope PSF.

In amateur setups - we have much more variable conditions - each sub has different FWHM and even if we dither, question is can we precisely do partial pixel shifts (we can't really and we rely on randomness).

Measure your FWHM to see how much you are undersampled, if any. With 4.51"/px - you need to be below well 7.2" FWHM in order to be under sampled.

[kentnek]

25 minutes ago, vlaiv said:

you need to be below well 7.2" FWHM in order to be under sampled.

Hmm I applied the FWHMEccentricity script to my master light's Luminance channel, and saw FWHM was around 2.440 px. Does this mean it's 2.440 × 4.51 ≈ 11 in arcsecond? 😨 

I wasn't too happy with this data as during integration I realized the sky was quite foggy, but I proceeded with stacking anyway to learn the process in PI. I guess drizzling will not help much then 🤣

25 minutes ago, vlaiv said:

even if we dither, question is can we precisely do partial pixel shifts

Agreed! I'm only dithering to reduce the walking noise on my Canon camera. Without guiding I have to nudge the declination and RA a little bit, even so it translates to quite a large displacement (dozens of pixels).

Edited January 10, 2022 by kentnek

[vlaiv]

3 minutes ago, kentnek said:

Hmm I applied the FWHMEccentricity script to my master light's Luminance channel, and saw FWHM was around 2.440 px. Does this mean it's 2.440 × 4.51 ≈ 11 in arcsecond?

Yes.

Did you use equipment in your signature?

Such high FWHM should not really surprise you, and it's not a bad thing. You are shooting wide field with that setup and low sampling rate is ok for such scenario.

You have 55mm of aperture and with that aperture alone - airy disk diameter is ~4.7". If you use any sort of field flattener - those usually produce PSF that is not diffraction limited. They fix up corners of the image but overall sharpness suffer. Add to that seeing and mount performance and you can easily get large FWHM even if your focus is spot on - and sometimes a bit of defocus also adds to the problem.

As far as I can tell - you have a bit of an issue with tilt in your system? Stars in the right part of the image are a bit astigmatic:

That also adds to the issue.

11 minutes ago, kentnek said:

I guess drizzling will not help much then

Yes, I'd avoid it if at all possible.

There are several tutorials out there that promote its use for some reason, and I often hear that people use it because they saw it in being used in tutorial.

In most (if not all) cases - there is really no need to use it.

[kentnek]

5 minutes ago, vlaiv said:

As far as I can tell - you have a bit of an issue with tilt in your system? Stars in the right part of the image are a bit astigmatic:

Yes, you're spot on! I do suspect there's a slight tilt in my imaging train, and attaching a full frame sensor to the scope magnifies the problem. However I'll be upgrading to an ASI533MC soon, so the tilt won't affect the 1" sensor that much I think.

13 minutes ago, vlaiv said:

In most (if not all) cases - there is really no need to use it.

Apart from the lower SNR on fainter nebula, I do see the drizzled version is cleaner. Would that be a benefit?

[vlaiv]

14 minutes ago, kentnek said:

Apart from the lower SNR on fainter nebula, I do see the drizzled version is cleaner. Would that be a benefit?

I think that is down to automatic stretch rather than anything else.

There is no reason why drizzled version would be cleaner as far as noise goes - on the contrary.

Here is what you can do to test things. First is - measure values on linear stack.

Make sure you register against same reference frame and that any scaling is the same for both stacks so (best to do it without scaling and work with actual ADU values or electron counts) and make selection of same part of background sky and do measurement - look at standard deviation value on same patch. Higher stdev means higher noise.

Another thing which is more geared to visual comparison would be - take above linear stacks and simply resize regular stack to match the size of drizzled stack (they will have different sampling rate so you need to up sample regular stack). Then take one half of up sampled regular stack and copy/paste it over drizzled stack - in the same place.

Now you have linear "split screen" to compare. What ever stretch you apply will be applied to both images and you'll have direct comparison between the two.