Compression by Noise Equalization

For any relation of the noise variance with the gray value, a nonlinear transform h(g) can be applied so that the variance of the transformed signal h is constant. The number of bits required to represent the noise-equalized signal is in good approximation equal to the maximum signal/noise ratio (SNRmax). Thus the noise-equalized signal of any imaging sensor with a full-well capacity of less than 2^16 can be represented by only 8 bit or less with only a slight increase in the overall noise level caused by additional quantization noise. The procedures are illustrated using the pco.edge camera./p>