Based Contourlets And An Image Of Total Variation Denoising Method

It is widely accepted that the wavelet transform is a very attractive tool to deal with non-stationary signal due to its multi-resolution and time-frequency location property. However, the approach is motivated by wavelet image denoising methods, where thresholding small wavelet coefficients leads pseudo-Gibbs artifacts. The total variation minimization can preserve sharp edge, but it can't preserve texture. In this paper, in order to preserve more texture while denoising, we use an efficient directional multiresolution image representation called contourlet transform, which is superiors compared to wavelets in capturing fine contours such as edge and texture. Unlike most conventional TV minimization techniques in image processing, the total variation regularity is directly imposed on the contourlet domain. The associated Euler-Lagrange equations lead to nonlinear partial differential equations(PDE's) in contourlet domain and proper numerical algorithms and schemes are designed to handle their computation. Our method performs a nearly artifact free image denoising, and achieves a better compromise between noise suppressing and edge preserving. Modifying our method, then we have a image inpainting method. The method is fit for inpainting images with geometric features. It is a worthy research on contourlets with its particular characteristic.