Noise Estimation and Type Identification in Natural Scene and Medical Images using Deep Learning Approaches

Table 6

Estimated noise level of five noise types.

Noise type	Added noise level	Estimated noise on natural images	Estimated noise on medical images
Gaussian	Mean = 0, variance = 2	280,380	2,687
Motion	Pixels = 15, angle = 20	2,195	1
Poisson	Variance = 0.04	34,311	4,740
Salt-and-pepper	Density = 0.05	71,220	46,740
Speckle	Variance = 0.04	54,126	9,203

The table shows the estimated noise levels for Gaussian is 2,687, motion noise is 1, Poisson noise is 4,740, salt-and-pepper noise is 46,740, and speckle noise is 9,203. From this, it is not possible to corelate the estimated noise level with the added noise.