Image Restoration is an important part of many image processing applications. Its main goal is to recover the original image form a degraded observation as shown in Fig. 1.
The existing linear image restoration algorithms assume that the PSF (point spread funtion) is known a priori and attempt to invert it. However for many situations PSF is not known explicitly and one has to estimate the true image and the PSF simultaneously using partial or no information about the imaging system, hence the process is called blind image restoration.
Among many approaches to blind deconvolution of images, I implemented and tested the iterative blind deconvolution (IBD) method and McCallum's simulated annealing algorithm. Those algorithms fall under the Nonparametric Deterministic Image Constraints Restoration Techniques. This class of algorithms do not assume any parametric models for either the image or the blur. Deterministic constraints such as nonnegativity, known finite support, and existance of blur invariant edges are assumed for the true image.
In addition to image restoration, the general blind deconvolution is applied to many other applications such as seismic data analysis, blind of communication channels, transmission monitoring, and echo cancellation in wireless telephony [3].
