Trimmed thresholding with SURE for denoising signals

We propose a signal denoising method based on the undecimated wavelet transform. The proposed method uses the well-known Stein`s Unbiased Risk Estimator (SURE). A derivation of SURE with trimmed thresholding of the undecimated wavelet transform coefficients is given and an adapted algorithm is developed to solve the resulting minimax problem. The result is a denoising method with better performance than the traditional thresholding techniques with soft thresholding. We illustrate the advantage of the derived wavelet denoising procedure on an ECG signal contaminated with background white Gaussian noise. The results show that the proposed method compares favorably with other methods.