An Efficient Denoising Method Using Dual-tree Wavelet Transform

Abstract

Denoising technology is an important operation in image processing. Denoising technologies approximately divides into two types: one is denoising operation in the spatial domain and the other is that in the frequency domain. Denoising operation in frequency domain is better than that the spatial domain in recently researches. The wavelet transform provides a multiresolution representation using a set of analyzing functions that are dilations and translations of a few functions (wavelets). The wavelet transform comes in several forms. The critically-sampled form of the wavelet transform provides the most compact representation, however, it has several limitations to denoise. Therefore, we describe dual-tree complex wavelet transform to solve these limitations. The image edge information maybe loses by denoising, so preserving edge is critical. We adopt the TV filter and Canny filter to preserve and enhance the edge information in low frequency subband. The simulation results demonstrate that that complex dual-tree method removes more noise signal than separable and real methods (all of the methods include both TV filter and Canny filter to preserve the edges). Dual-tree method outperforms separable method.