Parallel algorithms for restoration of degraded images using Gibbs field models.

摘要

In this thesis, restoration of noisy images using Markov Random Field (MRF) models for the clean images and the Maximum A Posteriori Probability (MAP) approach is considered. The degradation model is very general, in that it incorporates (a) additive and multiplicative random noise, (b) nonlinear distortion and (c) focus blur. Several algorithms are derived for restoration purposes, depending on the a priori information about the nature of degradations. These algorithms will allow for a very efficient implementation to obtain the restored images, also a general restoration algorithm has been developed for the case when the image has been corrupted by all the different components of the degradation. The algorithm was tested on a class of images representing: (1) Images ranging from binary (two levels) to continuous grey scale. (2) Degradation ranging from simple additive noise to the general degradation including additive/multiplicative noise, focus blur and nonlinear distortion. Test results indicate both the feasibility as well as the robustness of the algorithms. Finally, it is shown that the algorithms lend themselves to efficient hardware implementation, using SIMD/MIMD architecture as well as the more modern parallel architecture.Dept. of Electrical and Computer Engineering. Paper copy at Leddy Library: Theses u26 Major Papers - Basement, West Bldg. / Call Number: Thesis1989 .E543. Source: Dissertation Abstracts International, Volume: 50-03, Section: B, page: 1058. Thesis (Ph.D.)--University of Windsor (Canada), 1989.