Open AccessQuality measures in applications of image restoration
Author(s) -
Kriete Andres,
Naim Maria,
Schäfer Lutz
Publication year - 2001
Publication title -
scanning
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.359
H-Index - 47
eISSN - 1932-8745
pISSN - 0161-0457
DOI - 10.1002/sca.4950230504
Subject(s) - deconvolution , image (mathematics) , image restoration , computer science , logarithm , measure (data warehouse) , image quality , quality (philosophy) , set (abstract data type) , noise (video) , artificial intelligence , basis (linear algebra) , algorithm , image processing , computer vision , pattern recognition (psychology) , data mining , mathematics , mathematical analysis , philosophy , geometry , epistemology , programming language
We describe a new method for the estimation of image quality in image restoration applications. We demonstrate this technique on a simulated data set of fluorescent beads, in comparison with restoration by three different deconvolution methods. Both the number of iterations and a regularisation factor are varied to enforce changes in the resulting image quality. First, the data sets are directly compared by an accuracy measure. These values serve to validate the image quality descriptor, which is developed on the basis of optical information theory. This most general measure takes into account the spectral energies and the noise, weighted in a logarithmic fashion. It is demonstrated that this method is particularly helpful as a user‐oriented method to control the output of iterative image restorations and to eliminate the guesswork in choosing a suitable number of iterations.