Image denoising based on dual-tree complex discrete wavelet transform

Abstract

Complex discrete wavelet transform (CDWT) has significant advantages over real wavelet transform for certain signal processing problems. In this paper, denoising of SAR images based on Complex discrete wavelet transform (CDWT) was proposed. The dual-tree CDWT uses length-10 filters to obtain their real and imaginary parts. The top-level denoising strategy same as most other wavelet based denoising algorithms; decompose the noisy image into L levels of 6 directional subbands and a set of (lowpass) scaling coefficients, denoise the complex wavelet coefficients in each subband except for the scaling coefficients and invert the transform to obtain the image estimate. Experiment result shows a great effectiveness of Dual-Tree CDWT in removing the noise compared to the classical DWT.