Nonlinear filtering for image restoration

摘要

We consider a nonlinear filtering problem arising from the restoration of degraded images, in which both the true and the observed degraded images are described by nonlinear multidelay stochastic equations. The nonlinear filtering problem consists in computing the conditional probability density of the true image, based on the information generated by the observed degraded image. An explicit recursive formula for computing optimal filtering is obtained by applying the Girsanov transformation technique.