High ISO noise stretched into mid-brightness

[skygaze]

Hi folks, I've been trying to process my wide field image of Y4, and find that ST 1.5.369MR3 really cranks the high ISO noise. Canon 60Da, 60s subs x 36 ISO 6400 raw, with bias, flat, darks applied. Same result when stacked with Deep Sky Stacker 4.2.3 and Images Plus with differences only at the pixel level. In ST I've binned/cropped. Single frame histogram has the lowest fifth black and the peak at about 1/4 from the left. Yes, a little low... but see last paragraph.

Method_comparison.gif (478.61 KiB) Viewed 4756 times

Fundamentally I know there is a lot of noise at the bottom end. At far right just going straight from DSS into GIMP and using curves I can hide the worst of it, but I would love to use ST too get rid of more of it. I've tried a range of ignore detail, dark anomaly, grain size, with no luck. Gradients are gone like magic! In my limited understanding, here is my golf analogy: ST takes fairway grass and clips it to green height where the blades are all but invisible; and I've gone and given it a field of shrubs to work with. ST stretches the high noise into the mid levels, then spreads all that light into the purple mush.

Is there a workflow where I can get ST to leave the noise at the bottom of the stretch? Will it help to make my ROI the 20-30 pixels around the comet? Alternatively, anything smaller than a few-pixel star must be noise, especially if it changes colour so quickly. Am I missing some crucial settings?

I realize I can solve the problem by buying better equipment (longer exposures with lower ISO, faster lens, etc.) and my Star Adventurer appears to be a bit of a lemon, limiting me to 1 minute at 100mm focal length, but I'd prefer not to throw hundreds at it.
Regards.

[admin]

Hi and welcome to the ST forums!

Canon 60Da, 60s subs x 36 ISO 6400 raw, with bias, flat, darks applied.

First off, that ISO setting is way to high for your camera (see here). You need to greatly reduce it (try 800 for your model). ISO does not change sensitivity on digital cameras!

Will it help to make my ROI the 20-30 pixels around the comet?

Bingo! By default AutoDev will use the entire image as its Region of Interest (RoI). If there is mostly "nothing" (e.g. an "empty" background) in the image, then AutoDev will optimize the stretch to bring out that "nothing".
AutoDev just needs a good representation of what to bring out. If that is mostly "star like" objects, try an RoI around a star.
Right now though, it appears you have stretched the image far too much (going by the ST pre-noise reduction image).

Alternatively, anything smaller than a few-pixel star must be noise, especially if it changes colour so quickly. Am I missing some crucial settings?

Increasing the 'Ignore detail < parameter' will do exactly what you describe, e.g. it will make AutoDev "blind" to noise grain. The latter does not mean it will not show noise grain, it just means that it will no longer see it as "detail".
At any time, feel free to upload your stack if you'd like us to take a look!

Hope this helps & wishing you clear skies,

[skygaze]

I will try your suggestions on the ROI tomorrow (run out of personal discretionary time for today).

Regarding using a lower ISO, I appear to be painted into a corner. At 6400, here is the histogram for 60s:

Histogram_6400_60s.jpg (26.77 KiB) Viewed 4707 times

My earlier teaching said to push the ISO to the spot required to pull the histogram off the left edge (about 1/4). By going to ISO 800, it would be right against it. I "cannot" shoot 5 minute subs because my tracker isn't good enough. I did a bunch of sensitivity tests a year ago and found at 90 sec and 100mm I would lose more than half my frames to eccentric stars. Well, I will try a bunch at 800 and see what happens with the histogram way over. Like I said in my original post, my other options are a $2,000 f/2.8 lens, a $600 tracker, or shooting 5 minutes at 30mm. Sigh.

Hopefully the ROI around stars will work.

[admin]

Hi,

The histogram-rule-of-thumb is about figuring out the maximum exposure time you can use. You are certainly not required to meet that maximum if you cannot achieve it. Trying to achieve it by going beyond the native ISO of your camera is really not necessary or - the point of my advice - recommended. Does that make sense?

You certainly don't need "amazing" gear to achieve some great results (within reasonable expectations of course). There are many things you can do to optimize your gear and circumstances to get the most out of your gear! Throwing money at the hobby is not always (usually not!) the answer. Good calibration, acquisition techniques and, for example, dark sky sites will allow you to acquire very good, clean datasets. They may not be Hubble-quality, but good images of many brighter object should still be achievable.

[skygaze]

How about this?! Oodles better. Apart from using a very small RoI around the comet, I found the real key was to use gamma to push the low end darker. I think I used green cap a bit too much, because many of my stars have a small glacial lake blue core, and they somehow bloated a bit, but that's all fixable with more learning. Thanks Ivo!

20200322_Y4_DSS_STgam_heal_med.jpg (378.39 KiB) Viewed 4689 times

[admin]

That looks better indeed.
I don't seen any star bloat however... (usually by star bloat people mean the cores over exposing). The stellar profiles look nice and under control...

[skygaze]

So I tried a set at ISO 800 on T2 PanSTARRS, not Y4. Needed Siril to find stars because DSS got zero (too small and faint I guess). I can't say the result is markedly better , but the conditions (moonlight!) and total exposure times were worse, so that implies the result is better. Due to inbound cloud I could not run a second set at the different ISO setting to more readily compare apples to apples. Once again, setting gamma to 0.5 or less was key in suppressing the noisy background (this time due to low signal, instead of high noise). I've got some weird quasi-periodic mottling that I will need to try larger dark anomaly or grain or repair on.

So in short, thanks again Ivo for your help, this is much easier to control.
Alister.

20200403_T2_SirilScript_ST1_crop.jpg (134.79 KiB) Viewed 4632 times