Digital cameras contain light sensors which detect photons to produce a visual image using pixels of color. When pixels vary greatly from what a photographer knows the image to actually look like, the image is said to contain a lot of noise. This is similar to the "snow" effect seen on your television when the picture gets fuzzy. When you import a digital photo into Photoshop, the Adobe photo-editing program, you can select a portion of the photograph which appears fuzzy, and Photoshop will tell you the standard deviation of the signal (SNR for photography). This is the amount at which the pixels vary from the mean signal throughout the photograph.
Analyze is a program developed by the Mayo Foundation in Minnesota to calculate SNR from magnetic resonance imaging (MRI) machines. After an MRI captures an image using the single acquisition method, Analyze can determine SNR by using a series of mathematical equations. The single acquisition method requires only one MRI image, and it will compare the noise present in that image to a second image which is a blank control with no noise.
If the type of sensor and its properties are known by the operator, it is possible that SNR can be calculated manually through the use of equations. This also requires the signal types (audio, light, electromagnetic, etc.) to be known. These equations are complex, and may often require a number of variables to be known including wave frequencies, sensor dimensions, pulse durations and wave amplitudes. These equations are most useful under controlled laboratory conditions where these variables can be controlled. An example of an equation used for calculating the SNR in a nebula, according to the University of Vermont, is:
S/N = [Con - n(xsky)]/sqrt(Con-n(xsky))/G+n(sigma sky)^2 + n(sigma sky)^2/p
where:
S/N = Signal to Noise Ratio
Con = Counts by the sensor
n = star pixels
xsky = average counts by control sensor
sigma sky = RMS value of sensor
G = Gain
p = total pixels (star and sky)
Matlab is a computer software program that can determine the SNR without having to know what type of sensor was used. These parameters are inputted by the operator of the program. Matlab takes the data and reads it as being unitless. The operator gives the values units, but the program does not take this into account when analyzing. Matlab will output a numeric value which the operator must be prepared to interpret using the graphs Matlab produces.