What is Stochastic Resonance? - Alenas Imaging
Noise is usually considered the enemy of measurement precision - but in some circumstances, the right amount of noise can actually make an instrument more sensitive.
If many measurements are averaged, a vibrating ruler can measure more precisely than a fixed ruler.
Stochastic resonance or optimized dithering are signal processing techniques for increasing the precision of measurements when the measuring instrument is quantized. It depends on an optimum amount of noise and averaging of repeated, periodic data capture.
A simple analogy: Suppose you want to measure the length of a stick which is approximately 10 1/4 cm. long. But the only ruler you have available is marked at even centimeters only - nothing in between. This means that no matter how many times you repeat the measurement, you will find the stick is 10 cm long to the precision of the ruler, since that is the nearest mark.
Next suppose there is some noise or fluctuation present, causing the ruler to jitter or vibrate randomly. If the amplitude of the fluctuations is too small, it won't help much, and if they are too large, the measurement will be wiped out completely. But if the fluctuations are about equal to the 1 cm gap between quantized marks, and the measurement is repeated many times, then the stick will measure at the 10 cm mark about 75% of the time and the 11 cm mark about 25% of the time, for a statistical average of 10.25 cm. This optimum level of fluctuation for enhanced precision is the stochastic resonance.
In imaging, the CCD camera is the ruler and its grayscale levels are the quantized marks. With the optimum amount of noise and averaging of large numbers of images, increases in sensitivity up to 44dB are possible compared to a single image capture.
Stochastic resonance enhanced imaging can be applied to image very faint effects, such as those from thermoreflectance. Alenas calls this SRETR: Stochastic Resonance Enhanced Thermoreflectance.