The contribution deals with a new approach to the image analysis of complex
systems consisting of a large number of objects based on the neural networ
ks. It presents one original pre-processing method, which, together with an
appropriate neural network, gives successful results. While the new method
is tested on simulated hard-disk models, the results can be applied to exp
erimentally derived microphotographs of discontinuous and composite metal f
ilms. (C) 2001 Elsevier Science Ltd. All rights reserved.