A novel methodology for identifying noise types and estimating noise level was reported. The energy distribution of the HH subband's coefficients in the wavelet domain was used to distinguish Gaussian white noise and salt & pepper noise, which exist in images frequently. And the variance of the Gaussian white noise and the density of salt & pepper noise were estimated. The experiments on a wide variety of images indicate the veracity of distinguishing and estimating of noise.
Journal of Infrared and Millimeter Waves