Digital Image Restoration Using Conditional Markov Models.
Abstract
We are interested in developing digital image restoration algorithms using a class of spatial interaction models known as conditional Markov models. Our approach is to represent the images by Markov models on toroidal lattices and develop minimum mean square error (MMSE) restoration algorithms. The algorithms are non-recursive in structure and due to the underlying representation on toroidal lattices can be implemented using FFT algorithms. We give two types of algorithms. First we assume that a prototype of the original image is available and develop algorithms for restoration of degraded images. The degradation is assumed to be due to a known space invariant, non-separable, periodic point spread function and additive white noise. Secondly, we consider the more general situation when a prototype image is not available and give algorithms for MMSE filtering of noisy images. Experimental results are given for the above cases. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 01, 1981
- Accession Number
- ADA100002
Entities
People
- R. Chellappa
Organizations
- University of Maryland