Detector Blur and Correlated Noise Modeling for Digital Breast Tomosynthesis Reconstruction Supplementary Material

The derivation of the noise covariance of the ith projection view in Equation (5) is similar to the Appendix of Tilley et al. [22]; the difference is that our application is for DBT geometry in which the imaged volume is very close to the detector. As a result, we neglect the focal spot blur so the blurring operation in our study only includes the detector blur. For the directly measured projection before the log transform, we treat it as having a Gaussian distribution: ~ , , (32) where = exp ( − ), (33) = + , (34) where the capital letters Q and R in the superscripts denote that these values are for the projections before the log transform.