For distinction of the images having the identical geometrical sizes, as a sign distribution of spectral density of capacity of radiation on a surface of images is used. In the previous works as authors algorithms of distinction of images with the identical geometrical sizes have been developed at influence of compensated constant noise. However, in real conditions, along with them on an input of the receiver of optical radiation operate as well fluctuating noise, with a priori unknown level for each realization.
Article purpose is working out of optimum algorithm of distinction poisson and gaussian images with the identical geometrical sizes and definition of quantitative dependences of probabilities of a total error from characteristics of radiation of images and fluctuating noise.
On the basis of existing methods of synthesis of optimum intakes in the conditions of hindrances taking into account features of processing of signals in optical-electronic means are developed optimum (by criterion of the ideal observer) algorithms of distinction of images against fluctuating hindrances for cases continuous (Gauss) and discrete (Poisson) distributions of signals and hindrances. Researches of working capacity of algorithms are carried out and quantitative dependences of likelihood indicators of distinction on size of noise both for two-threshold distinction, and for a constant threshold are received. It is shown, that application of the two-threshold scheme in most cases provides lower levels of probabilities of a total error. It is established, that at equal values of noise in images distinction in signal levels in 5–10% does not render essential influence on likelihood indicators of distinction.
Xrebtov N.A., Malyarov V.G. Neoxlazhdaemy'e priemniki IK izlucheniya //
Opticheskij zhurnal. 1997.
T. 64. № 6.