Measuring noise in an image

[Robp]

Hi all,

Is there a way of measuring the amount of noise in an image?

I collect subs over multiple nights and I would like a way of measuring just how much of an improvement to noise each evening adds so I can tell if I should stop and move on to another target (diminishing returns) or just remove a certain nights images from the stack altogether?

[andrewluck]

You don't say which software you're using. Both Maxim DL and Pixinsight will do what you want and also have sub-frames selectors built to reject poor frames. I've no experience with other packages but most of them should have something similar.

Andrew

[Robp]

I primarily use DSS for stacking and Photoshop for processing 

[ollypenrice]

To halve the noise you need four times the exposure.

However, I don't think that a literal measuremnet of noise is really what we need. We can process selectively, so apply noise reduction (sparingly!) to such parts of the image as need it. We are not normally looking for high resolution in most faint parts of an image so the NR won't damage these regions. And our bright signal soon reaches the point of being as noiseless as it's going to get. The critical zones to look at are those in between. If you have moderately faint regions containing detail you want to stretch hard, or which carry detail, then these will define the signal strength the image needs.

In a nutshell, it's an image-by-image decision in my view. The same level of overall noise might be fine for a cluster but hopeless for a faint galactic tidal tail, for example.

Olly

[derrickf]

As Ollie pointed out in a rather long-winded way  the real measure you are looking for is Signal to Noise Ratio (SNR) rather than just noise. There is a ton of information about measuring , evaluating and optimising SNR in CCD imaging on the web but a good place to start to get a feel for how simple this is (NOT!!) can be found here

HTH

Derrick

[Robp]

Thanks for the replies both.

The main reason behind this is that I take around 10-12 hours of 20 minute exposures per filter on a target and as you say Olly, the brighter parts are fine but the 'middle bits' can be quite hard to tease out due to noise. I'm wondering if chasing after it is a lost cause due to my skies so I wanted a way to measure just how much the Signal to Noise Ratio improved every time I added a new nights worth of data.

Using the rule of four I'm assuming my milestones should be:

• 1x20 - 20 minutes total exposure time
• 4x20 - 1 hour 20 minutes total exposure time
• 16x20 - 5 Hours 20 minutes total exposure time
• 64x20 - 21 Hours 20 minutes total exposure time
• 256x20 - 85 Hours 20 minutes total exposure time

[Robp]

I've also found some information about how to measure the Signal to Noise Ratio in Photoshop. It is quite old (around 6 years) and on a rival forum so not sure if I can link it. Its by Craig Stark

You select a portion of background sky and then measure the mean and standard deviation of that selection. the Signal to Noise Ratio would be Mean/Standard Deviation.

When I get home I'm going to give that a try and see what I get.