Decomposition of an alignment problem of two 3D images by a multi-agent approach

In this work, we present an iterative technique for 3D images alignment based on multi-agent approach. This technique tries to reduce the dependency to the initial transformation from which suffers most of the traditional iterative methods. In deed, they require a good estimate of the transformation to guaranty the convergence to the correct solution. The proposed multi-agent approach considers the alignment problem of two 3D images not as a single problem, but as a set of alignment problems of the image regions. Then, a preliminary step of segmentation, followed by a region matching, is needed. The best solution is obtained by confrontation of the solutions given by each couple of homologous regions. This confrontation consists in applying two consecutive voting schemas. The first voting schema gives the best class of transformations, and the second gives the target solution.