The answer is "C", something entirely different.
The signal-to-noise ratio used in imaging is the ratio of the mean signal value divided by the standard deviation of the signal. This is used because photon counts, luminous intensity, etc., are always-positive values that are basically probability distributions.
Regarding dB (decibel) scale, imaging uses the field-quantity ratio (i.e., 20×log(ratio)) and not the power quantity ratio (10×log(ratio)).
So a 20 dB SNR means the signal-to-noise ratio is 10(20/20) = 10, which represents only log2(10) ≈ 3.3 bits of signal above noise floor. An 8-bit image with values equally distributed in the range 0 - 255 has an RMS value of about 147.4. Thus, the standard deviation among pixels values for a 20 dB sensor producing an 8-bit image is 147.4 / 10 = 14.74, or nearly 4 bits of noise. That is a very poor SNR for a sensor. A sensor with 20 dB sensitivity at a certain ISO sounds like either a very poor sensor, or a very high ISO, or a combination of both.