Shape Reconstruction in X-Ray Tomography from a Small Number of Projections Using Deformable Models

摘要: X-ray tomographic image reconstruction consists of determining an object function from its projections. In many applications such as nondestructive testing, we look for a fault region (air) in homogeneous, known background (metal). The problem then becomes the determination shape default region. Two approaches can be used: modeling binary Markov random field and estimating pixels image, or it directly this work model by deformable polygonal disc polyhedral volume propose new method coordinates vertices very limited number basic idea is not new, but other competing methods, general, modeled small parameters (polygonal shapes with vertices, snakes templates) these are estimated either least squares maximum likelihood methods. We polygon large allowing nearly any estimation vertices’ projections defining solution minimizer appropriate regularized criterion. This formulation also interpreted posteriori (MAP) estimate Bayesian framework. To optimize criterion use simulated annealing special purpose deterministic algorithm based on iterated conditional modes (ICM). results encouraging, especially when angles limited.