is drizzling worth it with a dslr?

[rustigsmed (Russell)]

hi,

since updating computers its allowed me more time to process images (soooooo much faster) stacking etc used to be around 40 mins now its done in under 5.

with all this spare time (and extra memory) I decided to play with the drizzle function in DSS. it only allowed certain sizes to use the function (custom rectangle). Is there a 'scientific' way to measure if there is any improvement to images (resolution)? my view it is perhaps a slightly better 'outcome' but that could be from processing the galaxies more tightly as they are done in crop rather than the actual drizzling .... hmmm.

thoughts?

Grus quartet "original size" (processed a few weeks back)
https://www.flickr.com/photos/803366...ream/lightbox/

Grus triplet 2x drizzle
https://www.flickr.com/photos/803366...ream/lightbox/

Grus triplet 3 x drizzle
https://www.flickr.com/photos/803366...ream/lightbox/

cheers,

rusty

[LightningNZ (Cam)]

Well to do a real comparison you need to keep the image scale the same. Can't really say anything else otherwise. I suspect it's not much worth it though as you're already close to oversampling with so many pixels in a modern DSLR.

[rustigsmed (Russell)]

Originally Posted by LightningNZ

Well to do a real comparison you need to keep the image scale the same. Can't really say anything else otherwise. I suspect it's not much worth it though as you're already close to oversampling with so many pixels in a modern DSLR.

hi cam,

that is a good point ... i might try and do a controlled size comparison.. as the seeing can be bad in Melbourne i was thinking perhaps drizzle could help off set that a bit. in good seeing probably no benefit whatsoever.

[PeterEde (Peter)]

I can't even get drizzle to work

[alpal]

Yes it's worth it on a super closeup as it will stop pixelation.

[Shiraz (Ray)]

you can measure any changes in FWHM (on unstretched data) to see if drizzle improved resolution (scaling by the drizzle upsample factor of course).

On the whole though, OSC images are mainly estimated data already, so it is difficult to see how drizzle will get you much super-resolution. However, as Allan points out, it is a useful mechanism for resampling.

[RobF (Rob)]

If you've got lots of good data and your set up is undersampled, then drizzle is definitely the go. Might be worth playing with CCDCalc and reading up around the web what pixel/resolution combos tend to give the best outcome. What sort of scope and FL are you shooting with?

[rustigsmed (Russell)]

Originally Posted by alpal

Yes it's worth it on a super closeup as it will stop pixelation.

thanks for your input Al - something i might consider on planetary or small deep space objects.

Originally Posted by Shiraz

you can measure any changes in FWHM (on unstretched data) to see if drizzle improved resolution (scaling by the drizzle upsample factor of course).

On the whole though, OSC images are mainly estimated data already, so it is difficult to see how drizzle will get you much super-resolution. However, as Allan points out, it is a useful mechanism for resampling.

Thanks Ray - I had a feeling you'd respond to this thread I might give this a go tonight just to see if there is any numerical difference. i guess cropping prior to drizzling and 'non drizzling' would be the go, stitching upsampled images could get a tad massive.

Originally Posted by RobF

If you've got lots of good data and your set up is undersampled, then drizzle is definitely the go. Might be worth playing with CCDCalc and reading up around the web what pixel/resolution combos tend to give the best outcome. What sort of scope and FL are you shooting with?

Hi Rob, thanks for the ideas, i haven't got CCDCalc so will look into it - my scope is 12" newt f4 (1200mm FL).

cheers

Rusty

[RobF (Rob)]

Here you go Russell

Calculates as about 0.74 arcsec/pixel.

You typically want resolution a bit under local seeing, so people tend to aim for around 1 arcsec/pixel as ideal. 2 arcsec/pixel and greater starting to get into "undersampled" range.

Conclusion - very little value from drizzle but huge extra processing overhead of already large DSLR files becoming 4 times larger.