Models and algorithms for edge-preserving image reconstruction

Image restoration and reconstruction are fundamental in image processing and computer vision has been reported. Although the physical processes involved and the numerical difficulties in implementing and executing the algorithms are very different in different cases, reconstruction and restoration are formally analogous and can be treated in a unified framework. The most common approaches proposed in the literature are based on MRF models for the image. Their success lies in the flexibility of MRF in modeling images with discontinuities and in introducing suitable constraints both in the image intensity and in the line field. Complexity is strictly connected to the nonconvexity of the posterior energy and to the usually very high number of variables to be estimated. The computation time required is still impractical for real-time applications when sequential computers are used. Algorithms that implicitly treat the discontinuities usually require less than 10 minutes even with sequential machines.