Premium
Three‐dimensional regularized binary image reconstruction from three two‐dimensional projections using a randomized ICM algorithm
Author(s) -
Retraint Florent,
Peyrin Françoise,
Dinten Jean Marc
Publication year - 1998
Publication title -
international journal of imaging systems and technology
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.359
H-Index - 47
eISSN - 1098-1098
pISSN - 0899-9457
DOI - 10.1002/(sici)1098-1098(1998)9:2/3<135::aid-ima11>3.0.co;2-w
Subject(s) - regularization (linguistics) , algorithm , binary number , simulated annealing , computer science , minification , iterative reconstruction , mathematics , mathematical optimization , artificial intelligence , arithmetic
In this article, we address the problem of fully three‐dimensional (3D) binary image reconstruction from three projections. The regularization approach relies on the use of a 3D Ising model which is a particular Gibbs prior. The problem is then equivalent to the minimization of a nonconvex energy function on discrete values. After an evaluation of standard optimization methods such as simulated annealing or its deterministic version ICM, we propose a new algorithm called parallel randomized ICM. It improves the reconstruction quality compared to ICM solutions while keeping reasonable reconstruction time. Its performances are evaluated from simulated projections for different 3D test images. Its application to real data is presented. © 1998 John Wiley & Sons, Inc. Int J Imaging Syst Technol, 9, 135–146, 1998