Abstract:

The purpose of this letter is to develop a new curvelet denoising algorithm for denoising images corrupted with additive white Gaussian noise (AWGN). We  used a total variation(TV) estimate as means to design a curveletdomain Wiener filter. The TV estimate indirectly yields the estimate of the image that is leveraged into the design of the filter. A peculiar aspect of this method is its use of TV and curvelet base, i.e., the TV for the design of the empirical Wiener filter and curvlet base for its application.  Numerical examples demonstrate that our method can perform better than curvelet shrinkage and TVbased method. Curvelets have been mathematically proven to represent distributed discontinuities such as edges better than traditional wavelets.

Key words: image denoising； curvelet； Wiener filter