Empirical Scatter Correction using the Epipolar Consistency Condition

Scatter affects every computed tomography (CT) image. Calibration-free software scatter reduction methods have not been used extensively in practice. Recently, consistency conditions have been applied successfully to other artifact reduction problems in CT imaging. We propose a scatter reduction method, that uses an epipolar consistency condition (ECC) to estimate parameters of an additive scatter model. We evaluate our approach by comparing it with an imagebased empirical scatter correction method (ESC) that uses the same scatter model. We show that it performs equally well on simulated data. Further, ECC outperforms ESC regarding the computational load for the determination of the parameter models, because ECC is formulated in projection domain such that no image reconstruction is necessary. While some restrictions might apply for the stability of ECC on measured data, no prior information needs to be formulated regarding the reconstructed image, like it is required with ESC.