Doctorant sous la direction de G. VALENZISE

Titre de la thèse: Fast and accurate 3D X ray image reconstruction for Non Destructive Testing industrial applications
Résumé de la thèse: 2D and 3D X-ray Computed Tomography is widely used in medical imaging as well as in Non Destructive Testing (NDT) for industrial applications. In order to reduce the X-ray dose in medical applications, and also for reducing the reconstruction time in NDT applications, we need to optimize the reconstruction methods while reducing the number of projections, thus using incomplete data to do the reconstruction. In the NDT applications, often the under-detected object is of huge size, which augments the difficulty of reconstruction. In our work we proposed to use the Bayesian Inference so that the prior information and measured data can be combined and various prior models could be applied, at the same time all the variables and parameters can be estimated directly. Mainly two methods are proposed for the piece-wise continuous image reconstruction problem. The first one reconstruct the object while enforcing sparsity of the haar transformation coefficients of object. The second method reconstruct objects while the contours are estimated simultaneously. In our work, all the proposed methods are adapted to the 3D huge data size case, while in 3D reconstruction application we uses the GPU processor to accelerate computation.