Abstract

Image restoration problems are commonly complicated by the presence of signal-dependent noise (SDN) sources (e.g., film grain noise, photoelectronic shot noise, speckle). As a consequence, both optimal and suboptimal restoration techniques are frequently nonlinear. This paper presents a review of research on techniques for using SDN models in image processing. Topics discussed include

(1) noise source modeling;

(2) nonlinear optimal/suboptimal restoration in SDN;

(3) image recovery from SDN only;

(4) adaptive point restoration;

(5) restoration using a Markovian covariance model;

(6) robust restoration techniques; and

(7) restoration based on an initial transformation of SDN to SIN (signal-independent noise). Both analytical results and the results of computer simulations on various test images are presented.

© 1986 Optical Society of America