Simulation and lossless compression of linear image sensor signals with more than 16 bit dynamic range

High-end scientific CMOS (sCMOS) sensors with a linear characteristic have reached already a quality level that 16-bit quantization is required in order to resolve the low temporal noise level in the dark image. Further progress will sooner or later require a quantization resolution of more then 16 bits.
This poster descibes that the signals of linear image sensors with high dynamic range and large maximum SNR, can be represented with a bit resolution far below 16 bit, keeping the processing fast. In addition, a non-linear gray value scale with gray-value independent temporal noise is advantageous for almost any type of image processing except radiometric evaluation. For the latter it is easy to linearize the signal again by applying the inverse gray value transform.