Level set methods for dynamic tomography

In this paper, we propose a novel variational framework for the reconstruction of dynamic objects from sparse and noisy tomographic data. Using an object-based scene model, we developed a general object dynamic model based on a one to one and differentiable mapping. We then propose a novel distance between curves to incorporate the object dynamics into the variational framework. For the minimization of the energy function, we developed a coordinate descent algorithm based on the level set methods. Experimental results for reconstructing a sequence of multiple dynamic objects are presented.