Wavelet and FFT based image denoising using non-linear filters

Abstract

We propose a stationary and discrete wavelet based image denoising scheme and an FFT based image denoising scheme to remove Gaussian noise. In the first approach, high subbands are added with each other and then soft thresholding is performed. The sum of low subbands is filtered with either piecewise linear (PWL) or Lagrange or spline interpolated PWL filter. In the second approach, FFT is employed on the noisy image and then low frequency and high frequency coefficients are separated with a specified cutoff frequency. Then the inverse of low frequency components is filtered with one of the PWL filters and the inverse of high frequency components is filtered with soft thresholding. The experimental results are compared with Liu and Liu's tensor-based diffusion model (TDM) approach.