ABSTRACT
In many imaging applications, the measured image is a degraded version of the true (or original) image that ideally represents the scene. The degradation may be due to (1) atmospheric distortions (including turbulence and aerosol scattering), (2) optical aberrations (such as diffraction and out-of-focus blur), (3) sensor blur (resulting from spatial averaging at photosites), (4) motion blur (resulting from camera shake or the movements of the objects in the scene), and (5) noise (such as shot noise and quantization). Image restoration algorithms aim to recover the true image from degraded measurements. This inverse problem is typically ill-posed, meaning that the solution does not satisfy at least one of the following: existence, uniqueness, or stability. Regularization techniques are often adopted to obtain a solution with desired properties, indicating a knowledge of prior information about the true image.
