Aiming at the shortcomings of traditional hard threshold method and soft threshold method, this paper puts forward a wavelet de-noising method based on wavelet entropy and improved threshold function. First, the noisy partial discharge (PD) signals are processed by wavelet decomposition, then the wavelet coefficients are processed by a new improved threshold function using adaptive selection of threshold based on wavelet entropy, finally the denoised signal can be got by reconstruction. The de-noising results of typical simulative signals and the field signals reveal that this present method can effectively remove white noise.