|Top Previous Next|
SensorNoise is used for two related but distinct purposes:
(1) spatial integration of sensor outputs (or actually, of any input Grid);
(2) addition of photoelectron conversion shot noise, other electronic noise and background variations, and A/D quantization.
Spatial integration is performed by specifying a new mesh whose spacing is coarser than the mesh on which the input exists. See the LightLike User Guide for details regarding the spatial integration rules and the mesh specifications. To only add noise without spatial integration, specify a new mesh that is identical to the input mesh. The new mesh defines the size of "physical" sensor pixels in which the subsequent noise effects and A/D quantization are computed.
The noise effects available are photoelectron conversion shot noise and signal-independent noise (representing, e.g., CCD read-out noise). The signal-independent noise in the physical pixels consists of independent samples of a Gaussian random variable with specified rms value. Additionally, nonuniform background and responsivity maps can be specified. Input quantum efficiency and responsivity (digital counts per photoelectron) are used to convert the input integratedIntensity to digital counts. Finally, a saturation value of digital counts can be specified.
All noise effects can be turned off by means of a single parameter, "addNoise"; this includes the truncation associated with A/D digitization. However, the conversion from input integratedIntensity units to counts units will always be performed: if "addNoise" = "false" the difference is that fractional counts will be reported.