Pose estimation of known objects during transmission tomographic image reconstruction

摘要: We address the problem of image formation in transmission tomography when metal objects known composition and shape, but unknown pose, are present scan subject. Using an alternating minimization (AM) algorithm, derived from a model which detected data viewed as Poisson-distributed photon counts, we seek to eliminate streaking artifacts commonly seen filtered back projection images containing high-contrast objects. show that this minimizes I-divergence (or equivalently, maximizes log-likelihood) between measured model-based estimates means data, converges much faster knowledge high-density materials (such brachytherapy applicators or prosthetic implants) is exploited. The algorithm incorporates steepest descent-based method find position orientation (collectively called pose) This pose then used constrain pixels their attenuation values, or, for example, form mask on "missing" shadow Results two-dimensional simulations shown paper. extension methods three dimensions outlined