On the determination of threshold in threshold-based de-noising by wavelet transform

The hard threshold t=cσ is proved to be efficient in threshold-based nonlinear filtering by wavelet transform. Generally speaking, the optimal parameter c varies with the signals and the intension of the added noise. Here we construct a function that approximates an equivalent function of the MSE of the recovered signal. The approximate optimal c can be determined by minimizing the function. Simulations show that the function we construct is consistent with equivalent function of the MSE of the filtered signal so that selected c is reasonable.