The Quincunx Filter-Based Sharp Frequency Localized Contourlet Transform for Image Denoising

Abstract

The contourlet transform, one of the recent geometrical transform offers the two important features of anisotropy scaling law and directionality, but non-ideal filters are used which results in two important problems, first is the pseudo-Gibbs phenomena around singularities produced by the Laplacian pyramid stage. Sharp frequency localized contourlet transform (SFLCT) is a new construction contourlet which succeeded in solving this problem by replacing the Laplacian pyramid with a new multiscale decomposition defined in the frequency domain. But the shift variant, the second problem was not solved by this work due to the downsampling of Laplacian pyramid and DFB stages. In this work, a high-performance quincunx filter banks are used in the DFB stage of SFLCT and noble identity is employed to solve the problem introduced by the downsampler and upsampler. The simulations illustrate significant improvements under the proposed transform when compared with other transform variations.