Affiliations: Department of Engineering Physics, Tsinghua
University, Beijing, China | Key Laboratory of Particle and Radiation Imaging,
Ministry of Education, Beijing, China | Biomedical Imaging Cluster, CBIS/BME, Rensselaer
Polytechnic Institute, New York, NY, USA | Department of Mathematics, University of California,
Los Angeles, CA, USA
Note: [] Corresponding author: Liang Li, Department of Engineering
Physics, Tsinghua University, Beijing 100084, China. E-mail:
lliang@tsinghua.edu.cn
Abstract: Different from the single-energy CT (SECT), multi-energy CT (MECT)
acquires projection data at different energy spectra, which makes that the MECT
has more sparsity among the data of separate energy and over energy. In order
to maximize utilization of all these sparse characteristics, this paper
proposed a new tensor PRISM model to consistently treat a priori knowledge of
the low rank, intensity and sparsity with the higher-dimensional tensor
technique. The priori knowledge of low rank corresponds to the stationary
background and similarity over the energy, and the intensity and sparsity
represents the rest of image features at single energy. Then, the
regularization and convex minimization problem was solved by tensor unfolding
and an extended tensor-based split-Bregman algorithm. Different from the
previous PRISM algorithm, the new algorithm mixed and treated different
constraints consistently. Numerical experiments have shown that our tensor
PRISM approach performs much better than the popular l_{1}
regularization algorithm in terms of image quality for MECT.
