Adnan Gavili Gilan - Sharif University of Technology
Hossein Shokri Ghadikolaee - Sharif University of Technology

when a thin paper is printed on both sides and we decide to scan it, we have a mixture of pictures from both sides. To separate these images we use scanned image from both sides. In this paper, we present three type of pre-processing as linear source separation. For a better separation, we use this fact that images rarely have edges in the same place and direction. Therefore, we use Non-Sub Sampled Wavelet Transform (NSWT) and Improved Non-Subsample Contourlet Transform (NSCT) de-noising for final separation. By using NSWT and NSC, we will achieve to very good separation results, since those strongly perform in representing edges. Experimental results show the presented methods have better separation performance.

Image separation, contourlet, non-subsample wavelet, non-subsample contourlet