Reviews of topical problems

Image restoration with minimum a priori information

P.K. Shternberg Astronomical Institute at the M.V. Lomonosov Moscow State University, poselok Nauchnyi, Crimea, Russian Federation

A consistent approach to the image restoration problem is presented, wich does not use Bayesian a priori information. Photon noise is taken into account. The unknown object is treated as a multidimensional set of parameters that have to be statistically estimated in an efficient way. The approach is based on an extended notion of feasible estimate (in the sense of information theory) and on Occam’s razor rule of choosing the simplest object which is consistent with the data. Occam’s rule is applied by transformation to principal components of the inverse (or maximum likelihood) estimate, which are generated by Fisher’s information matrix. The same approach can also be applied to various other inverse problems.

Fulltext pdf (958 KB)
Fulltext is also available at DOI: 10.1070/PU1995v038n02ABEH000068
PACS: 95.75.−z, 95.75.Pq (all)
DOI: 10.1070/PU1995v038n02ABEH000068
Citation: Terebizh V Yu "Image restoration with minimum a priori information" Phys. Usp. 38 137–167 (1995)
BibTexBibNote ® (generic)BibNote ® (RIS)MedlineRefWorks

Оригинал: Теребиж В Ю «Восстановление изображений при минимальной априорной информации» УФН 165 143–176 (1995); DOI: 10.3367/UFNr.0165.199502b.0143

References (129) Cited by (22) Similar articles (20)

© 1918–2023 Uspekhi Fizicheskikh Nauk
Email: Editorial office contacts About the journal Terms and conditions