Partial Discharge (PD) measurement is widely adopted for assessing the insulation conditions of high voltage (HV) equipment. Wavelet transformation (WT) is one the de-noising techniques to extract PD signals from a variety of environmental noises and interferences. In wavelet-based PD signal de-noising, mother wavelet selection is one of the major challenges. This paper proposes a novel level-based (LB) automatic mother wavelet selection scheme in conjunction with discrete WT (DWT) for PD signal de-noising. The de-noising results on both simulative and measured PD signals showed that the proposed scheme could successfully and consistently extract PD signals from noisy signals and retain their integrity.