Curvelet Domain Empirical Wiener Filter

The purpose of this letter is to develop a new curvelet denoising algorithm for denoising images corrupted with additive white Gaussian noise (AWGN).In this letter,we use a total variation(TV) estimate as means to design a curvelet-domain Wiener filter. The TV estimate indirectly yields an 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: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.