Neural networks for discrete tomography
Kees Joost Batenburg, Walter A. Kosters · 2005
Discrete tomography deals with the reconstruction of binary images from their projections in a small number of directions. In this paper we consider possible neural network approaches to this tomographic reconstruction problem. In particular we are interested in methods that can compute reconstructions in real-time and make ecient use of prior knowledge about the images, even when this knowledge is dicult to model by hand. We propose both a feed-forward back-propagation network method and a Hopeld network method for solving the reconstruction problem. 1