集装箱CT检测系统中贝叶斯图像重建算法的应用探讨

Bayesian reconstruction algorithms in the container CT inspection system

最大后验概率(Maximum a posteriori,MAP)图像重建算法是一种贝叶斯算法(Bayesian Re- construction,BR),可克服最大似然(Maximum Likelihood,ML)算法不能抑制噪声、重建时间较长的缺点。探讨了适合集装箱CT检测的两种MAP重建算法:加权最小二乘-最大后验概率-共轭梯度算法(weighted-least-squares MAP conjugate gradient,WLS-MAP-CG)和有序子集-贝叶斯算法(modified or- dered subset,MOS-BR)并讨论了合适的先验知识分布。

The use of maximum a posteriori （MAP） image reconstruction algorithms in the container CT inspection system is discussed here. The MAP algorithms are superior than the EM-based algorithms both in noise reduction and the convergence speed. We choose two efficient MAP algorithms： one is welghted-least-squares MAP conjugate gradient algorithms （WLS-MAP-CG）, the other is modified ordered subset Bayesian reconstruction （MOS-BR）. Accompanied with the Gibbs prior, both approaches can be expected to produce smooth reconstruction images with good quantitative accuracy.