Bin x2 for colour then drizzle x2. Does it make sense? - Forums

Topic starter

Jul 31, 2023 · Edited: Jul 31, 2023

Binning my CCD (KAF16200) for colour will improve the SNR, reduce exposure time, but at a cost of reduced resolution. In the same time I rely on the luminance data as the main driver for detail, so I won't expect much detail from the colour data anyways.
For bin x1 I am operating at a scale of 0.92"/pixels, with my summer average seeing around 2 -ish arc sec.

So if I colour bin x2 then do a drizzle x2 will the SNR return to the same value as doing bin x1 colour, or will I recover the resolution and retain a higher SNR.
To put it in other words, how much drizzling hurts the SNR …?

Engaging

dkamen

Jul 31, 2023

Hi,

How much drizzling affects the SNR is not a simple question to answer because it depends on droplet size, scaling factor, your exact sampling ratio, your read noise, how many subs you are taking and how they are dithered relative to one another. In general though, it is not drizzling itself that worsens SNR, it is the upscaling. In my experience drizzle usually scores much better than "naive" upscaling, but it is impossible to tell in advance.

However, I am not sure how relevant this all is to you because you are relying on L for resolution. AFAIK it is standard practice to just integrate colour in bin2x2 and then register the colour image with the luminance. Drizzling sounds like more trouble than worth.

Cheers,
D.

Helpful Insightful Respectful

AdrianC.

Topic starter

Jul 31, 2023

AFAIK it is standard practice to just integrate colour in bin2x2 and then register the colour image with the luminance. Drizzling sounds like more trouble than worth.

Cheers,
D.

Haven't tried it. So after I have my masters as rolled out from WBPP, what are the steps?

Christian Großmann

Aug 2, 2023 · Edited: Aug 2, 2023

Hi,

when I started astro photography, I put an ASI183 with really small pixels (2.4um) on a 10" f/5 Newton. With this kind of oversampling you have to bin your images. Otherwise your exposure times increase very badly. I can tell, because I tried

. So this is an example, where binning and drizzling make a lot of sense. At least for me…

CS

Concise Supportive

AdrianC.

Topic starter

Aug 2, 2023

Christian Großmann:
Hi,

when I started astro photography, I put an ASI183 with really small pixels (2.4um) on a 10" f/5 Newton. With this kind of oversampling you have to bin your images. Otherwise your exposure times increase very badly. I can tell, because I tried . So this is an example, where binning and drizzling make a lot of sense. At least for me...

CS

Yes in this case it makes a lot of sense. Having small pixels will considerably increase the exposure time especially for faint targets, since the light is spread over more pixels and the SNR drops. I don't know why the modern trend is to have smaller and smaller pixels. If you have to bin the CMOS, the read noise will double, and this negates the low read noise advantage.

Helpful Insightful Concise

Dave Rust

Aug 2, 2023 · Edited: Aug 2, 2023

Keep in mind that the color information on a striped camera sensor, by design, is already half the resolution of the lumenance image. So you would be cutting that in half again, only to use drizzle to approximately reconstruct the missing detail.

And the drizzle routine technically adds a bit of noise of its own. Most of us don't notice because the improved detail distracts us. So applying drizzle might only replace the noise reduced by BIN2. If so, the result would not be a better image at all.

Your suggestion might still lead to a subjective improvement, but I thought it was a good idea to remember this limitation going in. It will be interesting to find out what you learn.

Well Written Helpful Insightful Engaging

dkamen

Aug 3, 2023

AFAIK it is standard practice to just integrate colour in bin2x2 and then register the colour image with the luminance. Drizzling sounds like more trouble than worth.

Cheers,
D.

Haven't tried it. So after I have my masters as rolled out from WBPP, what are the steps?

Sorry for taking so long to reply, I was sick.

So basically you register all your masters with L (usually with a slightly cropped version of L since the images are unlikely to overlap 100%) and then process them normally. For L you do the full thing, DBE, deconvolution, stretching, masks and anything else you usually do. For RGB, just DBE, calibrate colors, stretch and maybe some denoising where you can be quite aggressive actually. Finally, LRGB combination with the stretched images. You will probably have to play a bit with the parameters as well as apply some finishing touches to the final result.

Example:

https://dktf.smugmug.com/Night/i-mGhRqN3/A

Cheers,
D

Helpful