The University of Edinburgh -
Division of Informatics
Forrest Hill & 80 South Bridge

Research Paper #624

Title:Stochastic Image Restoration: Clean Images and Their Likelihood
Authors:Perez-Minana,E; Fisher,RB; Wallace,D
Date: 1993
Presented:Presented at the Industrial and Engineering Applications of Artificial Intelligence and Expert Systems 6th International Conference, Edinburgh, 1993
Abstract:Geman and Geman (Geman 1984) make an analogy between images and statistical mechanical systems. The assignment of an energy function in the noise and the definition of an image restoration process based on: (i) the model described and (ii) techniques available for generating the maximum a posteriori distribution (MAP), using information from the corrupted image. The research described here analyzed the restoration of binary images obtained from the stochastic model by using the mean instead of the MAP estimate and concluded that equally satisfactory results were obtained at much less computational expense. In addition, it was shown that the process was not overly sensitive to parameter selection.

[Search These Pages] [DAI Home Page] [Comment]