The noise of a digital image consists of the signal noise and the camera noise which again comprises dark noise and readout noise. The dark noise is usually very low and comparable in size with the readout noise only in uncooled cameras. In case of Peltier-cooled cameras its contribution can be ignored even for very long exposure times. In general, the dominating noise factor in most experiments is the unavoidable signal noise; only for very low light intensities does the camera noise significantly influence the image quality.

Signal Noise (Photon Noise, Shot Noise): Light has an inherent noise that derives from the stochastic nature of the photon flux and equals the square root of the intensity. For a signal of 100 photons a noise of 10 photons or 10% results, while for 10,000 photons it is only 1% (100 photons). Since this noise is caused by the natural statistical variation of the light, it is not influenced by the camera design. Binning reduces the signal noise.

Dark Noise (Thermal Noise): This is caused by current fluctuations on the photodiodes in the absence of light arising from thermally generated electrons. The dark noise is the geometric mean of dark current and integration time. It is temperature dependent; this is why many CCD cameras are actively cooled, for example by Peltier elements. Roughly, a 8 – 10 °C decrease in temperature reduces the dark current by about half.

Read(out) Noise: This is the noise caused by the camera electronics, mostly by the on-chip preamplifier, upon quantification of the signal but also dependent on the A/D conversion rate. It is of significance only in low light applications.