Abstract
Tomography and its associated algorithms are now well known in the field of X-rays. On the other hand, all these tools are based on a modeling that differs from which could be envisaged in the field of Terahertz (THz) waves. We find, in the state of the art, models of propagation of the THz wave within an object. These models generate a THz wave that is either far from a ground truth, or of an algorithmic complexity that is too high to be used within a tomographic reconstruction in acceptable computing times. One of the objectives of this thesis work is therefore to obtain a propagation model of the THz wave allowing better modeling of the acquisition process and which can be calculated in relatively short times. When measuring the projection of an object, the absorption phenomenon is not the only phenomenon responsible for the attenuation of the THz wave. The phenomena of refraction and reflection are also responsible for attenuation of the measured THz wave. During a THz tomographic reconstruction, if these phenomena are not taken into account, the algorithm attributes this attenuation to the absorption phenomenon. This results in a reconstruction of the absorption coefficients of the object far from their real value. Under the effect of these phenomena, the problem of THz tomographic reconstruction is non-linear. This prevents the direct use of classical reconstruction methods since these methods imply that the relationship between an object and its projections is linear.