Neighboring Coefficients Preservation for Signal Denoising

For traditional thresholding denoising, the wavelet coefficients are thresholded without considering the information of other coefficients. In this paper, we propose a novel denoising approach which incorporates the neighboring coefficients into signal denoising. Our approach not only preserves the coefficients above the threshold, but it also preserves the coefficients predominated by useful components although their magnitudes are smaller than or equal to the threshold. Experimental results illustrate that the proposed approach is better than the NeighShrink scheme and the hard thresholding in both of the visual perception and the numerical results.