# Noise in Digital image

in Digital Image Processing restoration ,from my understanding after reading articles , it's seems that the noise must be dominant in the histogram of the image , which is make it possible , to get the mean and variance of the noise from the histogram it self.
My question is if I have a little Gaussian noise in the image , which i means that the noise is not dominant in the image , it's seems that it's wrong to extract the mean and variance from the image , because if we do this , we finding the mean and variance of the distorted image it self which is included the noise , yields that we affect on the original image if we use the model , which is included the model of the noise and the original image together, So how could we avoid this problem ?

• Compute the median. It is a robust mean. For the latter, use MAD (median absolute deviations), which is a robust deviation estimator. – Tolga Birdal Dec 30 '14 at 13:16