An MAP algorithm with edge-preserving prior for muon tomography

Muon tomography (MT) is an effective technology that is being developed to non-destructively detect high-Z materials. Because of the statistical randomness of scattering information and the data incompleteness, the widely used Maximum Likelihood (ML) estimation is often instable and accompanied with certain level of noise. In this paper, we propose a Maximum a Posteriori (MAP) algorithm with an edge-preserving prior to regularize the reconstruction. In each iteration the objective function is optimized using paraboloidal surrogate algorithm and the problem is transferred to solving a simple cubic equation. Experiments based on TUMUTY (Tsinghua University MUon Tomography facilitY) system demonstrate that MAP algorithm can effectively eliminate artifacts and performs well in local smoothing and edge preserving. Receiver operating characteristic (ROC) studies also verify this algorithm can significantly accelerate convergence and improve the reconstruction quality compared with ML algorithm.