Three dimensional (3D) images take an increasing importance in the medical imaging field. They may be produced by X-ray computed tomography (CT), magnetic resonance imaging (MRI) or position emitting tomography (PET) for example. They usually contain complementary informations which have to be combined together in order to be useful to physicians. Combining the information of sevral 3D images sets the problem of the registration of these images. It arises when two images of the same modality are taken with different positions or taken at different times (before and after a surgery for example) or with two images of different modalities. We present a new method for computing the rigid transformation between two 3D objets. Our method looks like a gradient method that minimizes the distance between both objects, but is more powerful because it is in fact an application of the fundamental laws of dynamics which uses both firt and second temporal derivatives of positions and allows an intuitive control of the minimization process. We show and discuss synthetic examples as well as real examples involving the registration of CT or MRI data of the head and the registration of PET and MRI images of the brain.