Abstract: A new de-noising algorithm is proposed. The characteristics of wavelet transform coefficients of general signal and white noise are analyzed. Hampel filter is modified for use in the de-noising algorithm. Taking advantage of the different characteristics of wavelet transform coefficients of white noise and useful signal, the modified filter is used to process the wavelet transform coefficients and reduce the noise in the observed signal. Experimental results show that the de-noising algorithm has good performance for different kinds of signals or different SNR (signal noise ratio), especially in the case of low SNR. So, it is a novel method for the detection of faint signals under a strong noise background.