# Should I increase the mean or the variance of Gaussian noise?

I am new in DSP.

I am currently working with a project and I want to try which algorithm works better given different image noises.

To test it out, I have been using the Gaussian noise in images.

I have been using this equation for imnoise:

f = imnoise(im,'Gaussian',0,0.001)


And then increasing by 0.001 and testing. As can be seen, I am using zero-mean noise. Is this the right approach to test the algorithm on different noises by changing the variance, or should I just change the mean?

• I would recommend you to change each parameter individually to understand its effect. This will let you get a feeling for the function and you will understand it better. In general if a function has multiple parameters it is highly recommendable to understand the reason/theory of the function. – Irreducible Apr 2 at 5:35
• Instead of changing the mean of the Gaussian noise, how about trying the following. Eliminate the Gaussian noise completely, and consider the effect on the image when you add just a constant to all the pixels in the image. What happens? Does the image look brighter? What happens if you change the constant from $0.01$ (say) to $-0.05$? How about to $0.15$? After looking at all this and observing the effects, put back the zero-mean Gaussian noise and look at the images again. This will show you how making the noise a nonzero-mean process will affect things and separate out the two effects. – Dilip Sarwate Apr 3 at 13:25