Obtaining the noise level of a camera

[Gifty01]

How can I subtract the camera noise from my real image? Basically, I want to calculate the center of gravity of a diffracted spot without the influence of my camera noise. I will be glad to know how this can be done.

 

[phyzguy]

Typically you take dark images without opening the shutter. Ideally you take a number of these, then the average of these is the dark signal and the standard deviation gives you an estimate of the noise.

 

[Drakkith]

How can I subtract the camera noise from my real image?

You can measure the noise, but you cannot subtract the noise. What you can subtract is the bias level of the camera sensor, the accumulated dark current, and anything else that adds a consistent unwanted signal to your image. Unfortunately the measuring and subtraction process inevitably adds additional noise to the final image, but with proper dark subtraction the amount of noise introduced in insignificant. The exact number of dark images you should takes depends on how much noise you are comfortable introduce. The more dark frames, the less noise.

 

[Twigg]

To give you a little math for this, the total signal on a given camera pixel might look like $$I = I_{signal} + I_{background}$$ $I_{signal}$ is the intensity of the image you're looking for, while $I_{background}$ is undesirable light that gets into your image. The noise on the pixel's intensity is given by the quadrature sum of signal shot noise, background shot noise, and detector noise: $$\sigma_I = \sqrt{I_{signal} + I_{background} + \sigma_{I,det}^2}$$
You can subtract out the background (you cannot subtract the noise) by taking another image in which you block the intended image and only measure the background. Then you have an image containing $I'=I_{background}$ only, with noise $\sigma_{I'}=\sqrt{I_{background} + \sigma_{I,det}^2}$. You then subtract these two images in software, and you get $$\Delta = I - I' = I_{signal}$$ with noise $$\sigma_\Delta = \sqrt{\sigma_I ^2 + \sigma_{I'}^2} = \sqrt{I_{signal} + 2I_{background} + 2\sigma_{I,det}^2}$$
The difference of images has only the intended signal on average, but is noisier than the original.

This doesn't tell you anything about the camera's noise. If you want to get rid of noise, take a boatload of images and average them.

 

[Gifty01]

You can measure the noise, but you cannot subtract the noise. What you can subtract is the bias level of the camera sensor, the accumulated dark current, and anything else that adds a consistent unwanted signal to your image. Unfortunately the measuring and subtraction process inevitably adds additional noise to the final image, but with proper dark subtraction the amount of noise introduced in insignificant. The exact number of dark images you should takes depends on how much noise you are comfortable introduce. The more dark frames, the less noise.

Thanks for your response

 

[Gifty01]

Typically you take dark images without opening the shutter. Ideally you take a number of these, then the average of these is the dark signal and the standard deviation gives you an estimate of the noise.

Thanks for your response. Is there a software in which I can just import the images and calculate the average as well as the std. Or I need to write a code? normally, I am not really good in coding. and I am newly learning how to use matlab. If it requires some code, I will be glad if an example can be sent to me. Thanks.

 

[Twigg]

filenames = {'image1.png','image2.png'}; %replace these with whatever image files you have
% make sure the image files are in the same directory as your MATLAB script

for n = 1:numel(filenames)
    image_stack(:,:,n) = imread(filenames{n});
end

image_average = mean(image_stack,3);
image_variance = var(image_stack,0,3);
image_deviation = image_variance.^(1/2);

figure(1); clf;
surf(image_average);
shading interp
view(2)

figure(2); clf;
surf(image_deviation);
shading interp
view(2)

That should do it. I can't test it for you because my personal PC is allergic to MATLAB since R2020a. Sorry

 

[Gifty01]

Thanks for your response. Kindly find attached the error that occured.

 

[Twigg]

Just replace line 5 with

image_stack(:,:,n) = double(imread(filenames{n}));

 

[Gifty01]

Thanks a lot twigg. I really appreciate.