A key parameter associated with imaging device performance is read noise, which varies for each detector. Even when shot noise is dominant, the read component of noise may be useful as a means of determining which detectors are faulty so that replacement values can be used. Furthermore, the aggregate rate of faulty detectors is useful as a figure-of-merit for the imaging device. Finally, the RMS variability of noise is a commonly used metric of overall imaging device quality. Typically, when raw detector outputs are analyzed in the presence of a stable background, the component of noise that is attributable to individual detectors is stationary, additive and uncorrelated between consecutive frames. Given a finite sequence consisting of N frames, the RMS sampling error in the estimate of the RMS noise σ is σ/sqrt(2N), when N is sufficiently large; this forms the basis of determining the needed value of N. Another consideration is the level of shot noise, since the read noise and the shot noise combine in quadrature. A third factor is the fidelity to which the read noise must be determined. This paper investigates the values of N that are needed for determining the detector noise with sufficient accuracy under varying conditions. Examples using simulated sensor data will be included.
|