Title

Image Restoration Based On the Fast Marching Method and Block Based Sampling

Author

Xiaoli Huan

Date of Award

2006

Degree Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Computing

First Advisor

Adel Ali

Advisor Department

Computing

Abstract

In this dissertation, a novel image restoration method is described, validated and compared with other available image restoration methods. The presented image restoration method combines the advantages of High Accuracy Fast Marching Method and non-parametric texture synthesis. The technique described from the method can handle both homogeneous and non-homogeneous natural images. The algorithm for the method can be elaborated in two sections. The first section of the algorithm is based on the High Accuracy Fast Marching Method to decide the filling order of the pixels in the target regions. The second section of the algorithm implicitly assumes a Markov random field model for textured image regions and blocks of texture are produced through an efficient searching process. The algorithm is easy to implement and requires only searching range of sample textures and a block size as user inputs. It restores the target regions with visually plausible quality equal to or better than those produced by previous techniques, with much less execution cost.