Deconvolution techniques for adaptive equalisation and image processing
Deconvolution techniques for adaptive equalisation and image processing
This thesis examines the application of recent advances in estimation theory to the problems of adaptive equalisation and image processing. In particular the algorithms described herein are based on Kalman filtering and dynamic programming techniques.It is well known that the Kalman filter provides the minimum mean square error (MMSE) solution to a wide range of filtering problems in which the data can be modelled as the output of a linear filter excited by a gaussian input. In data transmission systems the signal is generally binary so that improved performance at the removal of intersymbol interference can be obtained from non-linear processors exploiting this property and such filters are derived from modified Kalman equalisers and from simplified implementations of the Viterbi algorithm.In two dimensions the MMSE criterion is used to derive recursive algorithms based on Kalman filtering for the deconvolution of images blurred by camera/image motion or defocussing effects, and for optimum filtering of images heavily corrupted by white gaussian noise. Since the filters are applied in the space rather than the frequency domain they can be used for the removal of both spatially variant and invariant blurs.As an extension of the one-dimensional case, non-linear filters are derived for the processing of noise corrupted binary fields. A comprehensive collection of simulation results is presented demonstrating the performance of the algorithms.
University of Southampton
1976
Hart, Colin Graham
(1976)
Deconvolution techniques for adaptive equalisation and image processing.
University of Southampton, Doctoral Thesis.
Record type:
Thesis
(Doctoral)
Abstract
This thesis examines the application of recent advances in estimation theory to the problems of adaptive equalisation and image processing. In particular the algorithms described herein are based on Kalman filtering and dynamic programming techniques.It is well known that the Kalman filter provides the minimum mean square error (MMSE) solution to a wide range of filtering problems in which the data can be modelled as the output of a linear filter excited by a gaussian input. In data transmission systems the signal is generally binary so that improved performance at the removal of intersymbol interference can be obtained from non-linear processors exploiting this property and such filters are derived from modified Kalman equalisers and from simplified implementations of the Viterbi algorithm.In two dimensions the MMSE criterion is used to derive recursive algorithms based on Kalman filtering for the deconvolution of images blurred by camera/image motion or defocussing effects, and for optimum filtering of images heavily corrupted by white gaussian noise. Since the filters are applied in the space rather than the frequency domain they can be used for the removal of both spatially variant and invariant blurs.As an extension of the one-dimensional case, non-linear filters are derived for the processing of noise corrupted binary fields. A comprehensive collection of simulation results is presented demonstrating the performance of the algorithms.
This record has no associated files available for download.
More information
Published date: 1976
Identifiers
Local EPrints ID: 462636
URI: http://eprints.soton.ac.uk/id/eprint/462636
PURE UUID: 280dbad1-5108-4239-a391-d72f6ecdce7b
Catalogue record
Date deposited: 04 Jul 2022 19:34
Last modified: 04 Jul 2022 19:34
Export record
Contributors
Author:
Colin Graham Hart
Download statistics
Downloads from ePrints over the past year. Other digital versions may also be available to download e.g. from the publisher's website.
View more statistics