Maximum-probable restoration of photon-limited images

The paper presents a computer algorithm for obtaining physically maximum-probable (MAP) estimates of incoherent object scenes from their photon-count images. In the procedure, the Bayesian MAP estimation is combined with the known physics of image formation. Numerical tests of images in which data consisting of photon counts were generated by computer as independent random variables are presented. It is shown that the procedure results in the MAP object scene consistent with the image data; at the same time it allows the user to insert prior knowledge in a natural way, through a prior probability law.