Abstract
When an image or a document on a paper acquired through scanning, photographing or photocopying, the image on the back page may be show through. This cause a non-linear image mixture and should be resolved through a non-linear source separation. In this paper, we improve a previously introduced method, which suppose two independent images in the same point don't have same frequency. In this work we first give a mathematical analysis for an approximate linear separation as preprocessing; then nonlinear separation procedure is improved using Non-Subsampled Contourlet Transform (NSCT) instead of normal separable wavelet transform. The NSCT provides multiscale decomposition with directional filters at each scale. Furthermore, NSCT is very efficient in saving the geometric information of images and therefore it has very good feature localization. Experimental results show that our linear preprocessing and NSCT-based non-linear separation methods both have better separation quality, comparing with previously introduce methods.
Highlights
When we scan or photograph a paper document, especially if the paper is thin or transparent, the image from the back side often appears on the image from the front page
The coutourlet transform has the multiscale and time-frequency-localization properties of wavelets, and offers a high degree of directionality and anisotropy. This property of contourlet can be effectively used in our application, when we need more details in competition of components extracted in images scanned from each side of paper
The best separation were obtained with Non-Subsampled Contourlet Transform (NSCT) with proposed preprocessing for all mixed pairs, the result of using normal contourlet with proposed preprocessing is better than wavelet transform
Summary
When we scan or photograph a paper document, especially if the paper is thin or transparent, the image from the back side often appears on the image from the front page. Another work was based on the nonlinear denoising source separation (DSS) method [3]. This method suppose that two images have independent sources and have different frequency components in the same locations. Wavelet analysis offers limited directional information in representing image edges when separable one-dimensional transforms are used for images. Nonsubsampled wavelets are introduced for decomposing the image frequency components with better saving its geometric information.For comparing the results with [5], here we use the images that that Mr B. More information about images is available in [5] and [6].This paper is organized as follows: Section 2 explains the basic image separation method which is used in this paper.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have