A New Non-Aliasing Nonsubsampled Contourlet Transform
The factors of aliasing in the nonsubsampled Contourlet transform(NSCT) has been analyzed.The primary reason has been pointed that the à trous algorithm binary zero-interpolation brought about the width of the filter rapid increase and border distortion (that is aliasing). On that basis,a new approximate shift-invariant non-aliasing pyramidal decomposition was proposed instead of the à trous algorithm nonsubsampled pyramidal decomposition in the NSCT,So a new approximate shift-invariant non-aliasing nonsubsampled Contourlet transform(NANSCT) was constructed. Compared to the NSCT,the basis image of the NANSCT has better spatial domain regularity, frequency domain localization and decreased redundancy.The experimental results show that whether PSNR index or in visual effect, the proposed scheme outperforms the traditional Contourlet transform hard threshold denoising , Contourlet domain HMT denoising and the NSCT hard threshold denoising, and can achieve an excellent balance between suppressing noise effectively and preserving as many image details and edges as possiblet.
H. Yan and Y. J. Pan, "A New Non-Aliasing Nonsubsampled Contourlet Transform", Key Engineering Materials, Vols. 480-481, pp. 893-898, 2011