Projection data used for the reconstruction of low-dose Computed Tomography (CT) are degraded by many factors and pre-reconstruction corrections. These make the noise property of proj ection image difficult to analyze and render a very challenging task for noise reduction. In this study, we first investigate the nonlinear noise property of low-dose CT projection data by analyzing a repeatedly obtained experimental data set. Since the noisy CT projection data can be regarded as normally distributed, with nonlinear signal-dependent variance, we propose a K-L domain penalized weighted least-square (PWLS) smoothing method for the accurate treatment of this kind of noise.
展开▼