Noise reduction techniques

[Pryce]

I'm completely new to imaging, so I have many hurdles to overcome.

What I'm currently pondering over is noise.

What can I do to reduce this kind of noise?

I'm using a DSLR, so I understand that I am somewhat limited. But I hope theres some tricks I can use to get around it.

The image above is from a stack with 3h45m integration time. Do I need more subs? More darks or bias? Are therw stacking parameters that might affect this? Other techniques?

My equipment is listed in the signature and I use DSS and photoshop for post processing. I have thw Astronomy tools plugin for PS as well. 

[bottletopburly]

Can’t see your signature but dithering with a dslr is beneficial to eliminate walking noise ,I use Apt this communicates with phd2, more subs the better will help too .

[almcl]

Cooling a DSLR can help with noise, although how to achieve this is not always simple. 

But without a stable temperature on both lights and darks, calibrating with mis-matched dark frames can actually make matters worse.  Especially so when the temperatures don't match. 

Processing software too, can help with noise.  I've used a couple of modules from StarTools on your screenshot to quickly get this result:

[pete_l]

One comment about the original image. The stars seem to be larger than the noise pixels. They also appear redder at the top and bluer at the bottom.
Is this a focus issue - it doesn't look like it would need much of an adjustment, or maybe an optical one? Depending on where that cropped example was in the overalll image - centre, corner, etc. it could be that getting the star size smaller would make them brighter, so the signal (the stars) to noise (the background) ratio would automatically improve.

[Pryce]

6 hours ago, bottletopburly said:

dithering with a dslr is beneficial to eliminate walking noise

I dither 5px between every frame!
 

6 hours ago, almcl said:

calibrating with mis-matched dark frames can actually make matters worse.  Especially so when the temperatures don't match. 

Should I then calculate the median temperature of my subs and use the dark frames that are closest in temperature to that.  Or should II just deslect the outliers?
I will give startools a look!

 

5 hours ago, pete_l said:

The stars seem to be larger than the noise pixels. They also appear redder at the top and bluer at the bottom.

This was my first session, so nothing was perfect.  PA was probably off, focus was probably off(though I used a bahtinov mask), guiding didn't work, so I had to track with sidereal.  The cropped section was from the top left corner.
 

[bottletopburly]

I you dither 10-12 pixels  then Darks not required just bias and flats ,bias you can do every now and again three or four times of year and flats on the night,time saver .

Edited December 26, 2020 by bottletopburly

[Pryce]

Really?  So me dithering and using darks may have made the nosie worse?

 

Edited December 26, 2020 by Pryce

[Seelive]

You don't say how dark your sky is. If you suffer from light pollution (bortle 5+ ?) then at non-tropical night time temperatures the background sky noise is more than likely to be higher that the noise introduced by most modern DSLRs so darks won't add any real value.

[pete_l]

1 hour ago, Pryce said:

This was my first session, so nothing was perfect.  PA was probably off, focus was probably off(though I used a bahtinov mask), guiding didn't work, so I had to track with sidereal.  The cropped section was from the top left corner.

In that case I wouldn't worry. Everyone gains experience over time provided they keep practicing

So the next session will be better and so on.

[Pryce]

1 hour ago, Seelive said:

You don't say how dark your sky is. If you suffer from light pollution (bortle 5+ ?) then at non-tropical night time temperatures the background sky noise is more than likely to be higher that the noise introduced by most modern DSLRs so darks won't add any real value.

At this site, I'd assume either bortle 5 or 6!

So realistically, I won't be able to reduce noise any further by adding calibration frames, but instead I should focus on dithering, guiding and capturing more data to stack to increase the SNR? Possibly reduce the ISO for brigther objects like m42? (Was 800 in this picture)

[almcl]

2 hours ago, Pryce said:

Should I then calculate the median temperature of my subs and use the dark frames that are closest in temperature to that.  Or should II just deslect the outliers?

It might be worth reading the EXIF temperature data for your subs.  

If you don't already have an AP to do that, this one maybe of use:

http://www.covingtoninnovations.com/dslr/exiflog.html

You can see what temp is recorded for the lights and darks and if they match, try stacking both with and without darks and seeing which result you prefer?

 

If you want to go into it a bit more there's Dark Master which will automatically generate stacking lists of matched darks and lights, I tried using it but my subs were too variable for it to be much use.  Once I started using an external cooler on my 700d it wasn't really necessary.

Edited December 26, 2020 by almcl

[Pryce]

I'm using a canon, so the temperature actually get's placed in the filename

But yeah, I'm trying different stacking parameters now to see what works best!

[ollypenrice]

Let me give some pointers on post-processing noise reduction in Photoshop.

NEVER apply noise reduction globally. You have great selection tools, so exploit them. Use the Colour Select tool to pick regions of low signal (background sky and faint nebulosity) where the noise is at its worst and explore all the possibilities. Reduce colour saturation. Noise- Reduce Noise. Experiment with the sliders.

Zoom in to pixel scale and look for rogue pixels, some of which can be systematic. You might have a smattering of, say, rogue magenta pixels. Use colour select to select them and then experiment with the noise-median filter. You can usually make them vanish.

Try using layers: make a copy layer and use heavy noise reduction on the bottom layer. Back on the top layer use the selection tools to pick the noisy pixels and then use the eraser to remove them. You choose the ones to remove so this is non-global. Global is always bad.

I think that Photoshop offers the best noise reduction in the business precisely because it can be finely customised.

Olly

[Pryce]

These are great suggestions and I will definitely put them into use! 

 

 

[Pryce]

I applied some of your suggestions to my first DSO photo.
This first one was my WIP before applying your suggestions and where the screenshot from the OP was taken from.
The second one is my WIP after applying your suggestions.  Or at least some of them.

I may have gone a bit overboard though.  And I tried to mask out the stars and nebulosity, but seing as I probably had a focusing and/or tracking issue I think my star masks were a bit off.

Though I still feel that the image looks better in terms of noise. Coments concerns?

[ollypenrice]

That's good. I would now use the colour sampler Eyedropper to measure the background. I think you'll find the green channel a little high.

Olly