In the field of X-Ray Radiography, different three-dimensional reconstruction techniques have been proposed : apart from computer-aided tomodensitometric methods one can reconstruct an object by using the different slices which are obtained by the now classic scanner. Nevertheless, it is also possible to consider that problem more directly, by using a set of planar projections. In the parallel-beam geometry a set of projections obtained by the rotation of the beam in a plane is sufficient toireconstruct the 3 D object. In this case one can generalize the filtering and spreading process, usually carried out on the monodimensional projections, at the case of planar images corresponding to 2 D projections. However, this method meets technological difficulties as it is difficult to have a parallel-beam with a large cross-section. This fact has led us to consider the use of a conical beam. The adaptation of the previous results at this conical geometry which requires a conical spreading and a specific filtering, also makes necessary to get projections which have to be uniformly distributed in all orientations of the space. This causes further technological and practical difficulties due to the generally limited possibilities in the positioning and the displacement of the X-Ray sources. We are thus inevitably lent to use the measurements originating in a system based on the cone-beam geometry whose axis rotates only in one plane. Thus, this axial symmetry will cause a certain number of problems which are in fact, the consequence of,the limited knowledge of the information. The following, then, are the possible solutions : 1) To design on to use the acquisition system in such geometrical conditions that it is possible to disregard the effect of beam divergence. The results will then be identical to those used in parallel geometry 3 D reconstruction. 2) To use, in this particular axial symmetry, the same method of reconstruction as that considered when we have the whole set of conical projections of all directions in the space. In this solution, some errors appear owing to the lack of a large number of directions. 3) To develop algebraic methods of reconstruction adapted to the conical geometry, using some criteria of optimization. The latter solution seems to be the best approach but necessitates strongly in practice, the use of specific computers (like "array processors") in order to reduce the presently too long time taken by the computer process.