A new approach for low dose CT reconstruction with highly sparse projections

In this paper, we designed a low dose scan plan by reducing the projection angle of views. A Bayesian maximum a posteriori (MAP) method for corresponding reconstruction. L0-normprior is brought into our proposed method. To ensure the convergence of the algorithm, a set of surrogate potential functions are used to successively approximate the L0-norm function while generating the prior. To accelerate the convergence speed, SPS algorithm is applied. Since the projection is assumed to be with Gaussian statistics, the computational comparatively easy. Simulation results have proved that with the L0-norm prior, the proposed method is capable of providing exact reconstructions with highly sparse sampled noise free projections, even with noise contained projections, the reconstruction quality is still significantly superior to reconstructions with L1-norm or L2-norm prior.