Image restoration using a feed-forward error backpropagation neural network ensemble

A mathematical model is used to generate gray scale Images. This model Is based in the use of the Convolution theorem and of the Zernike polynomials. The model is built into a FORTRAN program so that it is possible to represent gray scale images distorted \with spherical aberration (as well as images with no aberrations). A database is constructed in such a way that pairs of aberrated and non-aberrated images are classified. This database is used train an ensemble of Backpropagation neural networks to correct this optical fault. The objective is to find a method to correct spherically aberrated images outside the original training database. This work shows promising results for the future application of similar techniques, to problems such as those found in the Hubble Space Telescope or other optical systems with similar optical aberrations,