Drizzle widefield image - Forums

Topic starter

Dec 28, 2023 · Edited: Dec 28, 2023

Hi
I am working on a widefield image of IC1805 and am aiming towards 60+ hours of narrowband data.

The lens used is the Samyang 135mm and an ASI2600MM pro - with that, I have a pixel scale of about 6 arcsec/pixel, which is significantly undersampled.
Yesterday, I tried to drizzle my data, which seems to have worked pretty well. I now have a pixel scale at about 3 arcsec/pixel and a much more pleasing image to look at.
Left is non-drizzled - right is drizzled - it's Melotte 15 heavily zoomed

What might be the downside to this approach?

I don't care about the larger images or the potentially longer process time, which to me is insignificant and a price I'm willing to pay if the end result as I currently see it seems much better.

Helpful Engaging

Jan Erik Vallestad

Dec 28, 2023

The price you pay is the increase in noise as you increase your image size, but you could also just resample the image afterwards to negate that effect, or try drizzle x1 which will not affect SNR. If you have lots of data maybe the change in SNR won't be that noticeable though.

Christian Bennich

Topic starter

Dec 28, 2023

The price you pay is the increase in noise as you increase your image size, but you could also just resample the image afterwards to negate that effect, or try drizzle x1 which will not affect SNR. If you have lots of data maybe the change in SNR won't be that noticeable though.

I guess that I could measure the noise in both the drizzled and non-drizzled images to see how much extra noise I get from the drizzle process?

Well Written Engaging

Brian Puhl

IOTD Staff

Dec 28, 2023

What settings did you use for drizzle? I'd expect with a roki you could see better results than that.. but I think some it is your process.

Either way as I advocate for drizzle in any situation where you're undersampled, the only downside is less SNR. I do not find more noise in drizzled images, simply less signal. Pump lots of integration time into your images and drizzle is easily your best friend for the Roki. Your stars will be smoother, no longer limited to single pixels, and the combination of BlurX in the mix will make your images appear far better sampled then they are. Now, if you're shooting after faint stuff, I'd keep that drizzle off, you're only hurting yourself. Bright targets with bold filament structure like Veil Nebula tend to really benefit the best from drizzle.

Helpful Insightful Engaging Supportive

Christian Bennich

Topic starter

Dec 28, 2023 · Edited: Dec 28, 2023

These are the settings I used for the Drizzle Integration.
Anything I could benefit from changing?

Brian Puhl

IOTD Staff

Dec 28, 2023 · Edited: Dec 28, 2023

Christian Bennich:
These are the settings I used for the Drizzle Integration.
Anything I could benefit from changing?

My best results so far have been with Var Shape 1.5 kernel, with drop shrink anywhere between 0.5 and 0.8 depending on how much integration time I have. Experiment with it, not every situation is the same. Compare them side by side and look for empty pixels when you've gone too far. If you want an example of just how well it can work, look at my Veil images in my gallery from a few months ago. Those settings worked almost too good lol.

Also, things could probably go a little faster if you turn off local norm.

Helpful Concise

Christian Bennich

Topic starter

Dec 28, 2023

Brian Puhl:
Christian Bennich:
These are the settings I used for the Drizzle Integration.
Anything I could benefit from changing?

My best results so far have been with Var Shape 1.5 kernel, with drop shrink anywhere between 0.5 and 0.8 depending on how much integration time I have. Experiment with it, not every situation is the same. Compare them side by side and look for empty pixels when you've gone too far. If you want an example of just how well it can work, look at my Veil images in my gallery from a few months ago. Those settings worked almost too good lol.

Also, things could probably go a little faster if you turn off local norm.

Thx, will test later tonight 👍

Mau_Bard

Dec 28, 2023 · Edited: Dec 28, 2023

Christian Bennich:
Hi
I am working on a widefield image of IC1805 and am aiming towards 60+ hours of narrowband data.

The lens used is the Samyang 135mm and an ASI2600MM pro - with that, I have a pixel scale of about 6 arcsec/pixel, which is significantly undersampled.

(...)

What might be the downside to this approach?

I don't care about the larger images or the potentially longer process time, which to me is insignificant and a price I'm willing to pay if the end result as I currently see it seems much better.

Hi Christian, IMO the Samyang 135mm + ASI 2600 combination is fabulous, yet definitively under-sampled; therefore drizzling is a must, in particular if you do not want to land with square stars after deconvolution. The f/2 gathers so much light that decreased SNR should rarely be an issue, and, in any case, you will gather by design enough exposures to sustain drizzling.
To me the only annoyance is that processing time becomes significantly longer than usual, because PI and the Xterminator suite do not use the GPU integrated in normal laptops.

Happy new year!
Mau

Helpful Insightful Respectful Engaging

Christian Bennich

Topic starter

Dec 28, 2023

Christian Bennich:
Hi
I am working on a widefield image of IC1805 and am aiming towards 60+ hours of narrowband data.

The lens used is the Samyang 135mm and an ASI2600MM pro - with that, I have a pixel scale of about 6 arcsec/pixel, which is significantly undersampled.

(...)

What might be the downside to this approach?

I don't care about the larger images or the potentially longer process time, which to me is insignificant and a price I'm willing to pay if the end result as I currently see it seems much better.

Hi Christian, IMO the Samyang 135mm + ASI 2600 combination is fabulous, yet definitively under-sampled; therefore drizzling is a must, in particular if you do not want to land with square stars after deconvolution. The f/2 gathers so much light that increased SNR should rarely be an issue, and, in any case, you will gather by design enough exposures to sustain drizzling.
To me the only annoyance is that processing time becomes significantly longer than usual, because PI and the Xterminator suite do not use the GPU integrated in normal laptops.

Happy new year!
Mau

Thanks, that is great to hear.

I am not suffering too much in regards to performance as I’m on an M2 😈😈
But it will be really awesome once the PI folks get the new release out that runs natively on OSX.

Happy new year to you as well and clear skies.

Willem Jan Drijfhout

Dec 28, 2023

Left is non-drizzled - right is drizzled - it's Melotte 15 heavily zoomed

Christian, the drizzled image looks indeed much better. You seem to have lost a lot of the weaker stars. Is that only drizzle effect, or more processing steps involved?

Well Written Respectful Engaging

Christian Bennich

Topic starter

Dec 28, 2023

Willem Jan Drijfhout:
Left is non-drizzled - right is drizzled - it's Melotte 15 heavily zoomed

Christian, the drizzled image looks indeed much better. You seem to have lost a lot of the weaker stars. Is that only drizzle effect, or more processing steps involved?

@Willem Jan Drijfhout - the “loss” of stars is intentionel.
I am using Bill Blanshans “Starreduction” script for that.
Personally I like the way it emphasizes the nebulosity when the stars are not that dominant.

Christian Bennich

Topic starter

Dec 28, 2023

@Brian Puhl - I reran the DrizzleIntegration now.
Here are the results on the stacked HA image.

Comparing 3 images.
Without Drizzle, With default Drizzle settings and finally with the settings you suggested.
Only thing I have done to the image is BlurXterminator - ONLY to see a simple result.
Default STF stretch.

TL;DR - The settings you provided make for yet another significant improvement to my image.
I am no master in understanding the SNR - as I interpret it - I get a significant improvement in my overall image, at the cost of a slightly higher noise level.

Overall - I think that we have a winner here.
Samyang 135mm + ASI2600 + Drizzle == 😍😍😍

Default Drizzle Setting:

New Drizzle Setting:

BlurXTerminator Setting:

Images 1 - NO Drizzle

SNR calculation in image:

Image 2 - Default Drizzle settings

SNR calculation in image

Image 3 - New Drizzle settings

SNR calculation in image

Helpful Engaging

Brian Puhl

IOTD Staff

Dec 28, 2023

It's hard to tell with an STF, (i'd relax that a bit cuz she's bright) but it almost looks like you ended up with a bit of a grid pattern in your third iteration. I guess you're at 2:1 scale though. Definitely a solid improvement. Like I said, it's alot of trial and error, but I'd say you're off to a really great start.

Christian Bennich

Topic starter

Dec 28, 2023

Brian Puhl:
It's hard to tell with an STF, (i'd relax that a bit cuz she's bright) but it almost looks like you ended up with a bit of a grid pattern in your third iteration. I guess you're at 2:1 scale though. Definitely a solid improvement. Like I said, it's alot of trial and error, but I'd say you're off to a really great start.

Yep - a bit of grid pattern in the there. Have to figure out how to get rid of that.

Brian Puhl

IOTD Staff

Dec 28, 2023

Christian Bennich:
Brian Puhl:
It's hard to tell with an STF, (i'd relax that a bit cuz she's bright) but it almost looks like you ended up with a bit of a grid pattern in your third iteration. I guess you're at 2:1 scale though. Definitely a solid improvement. Like I said, it's alot of trial and error, but I'd say you're off to a really great start.

Yep - a bit of grid pattern in the there. Have to figure out how to get rid of that.

That's kinda what I was getting at with the empty pixels part. Might have gone a tad too far due to lack of integration. but if you don't see it at a 1:1 scale, I wouldn't be concerned.

Christian Bennich

Topic starter

Dec 29, 2023

The faint grid pattern completely disappeared during processing.
Especially NoiseXterminator fixed it entirely as the last step in my process.