We present a new filtering method based on wavelet analysis. This method works without any assumption about the noise, and we show that it gives reliable results on X--ray spectroscopic data. The originality of this method consist in the wavelet analysis: the filtering is done both in frequency space and in the spatial space, thus avoiding artifacts due to fourier filtering. We also show the reliability of the method using test signals. All the software developped following this method is freely available on the network.