Multiple Model Kalman Filter Approach for Show-through Cancellation

Abstract

Documents of historical importance and other duplex printed ones often suffer degradation due to ink on one side seeping into the other side and/or contents on one side appearing and interfering with contents on the other side. These are respectively called as bleed-through and show-through. A dual estimation approach is proposed in the paper that uses both the recto side of the document and the verso side for show-through cancellation. The main assumption is that the degradation is because of the linear mixing of the recto (verso) image and horizontally flipped verso (recto) image. The state-space model framed to define the show-through can be used to estimate the states which are clear recto and verso images. The proposed method uses multiple Kalman filters to solve for state estimation and deal with unknown parameters in the model. The simulation results for different cases establish the validity of the proposed approach.