Reassigned spectrogram can effectively enhance the time-frequency concentration and reduce the misleading interference terms, however, the readability of time-frequency distribution of the reassigned spectrogram is reduced when there exists strong background noise in a signal. So, a new method for mechanical fault diagnosis based on reassigned Wigner-Ville distribution spectrogram (RWVDS) and singular value decomposition (SVD) is proposed. The RWVDS is obtained by using the reassigned algorithm to the Wigner-Ville distribution Spectrogram (WVDS). Then, SVD de-noising is applied to the RWVDS to improve the readability. The experimental analysis in the fault signal of simulation signal, rolling bearing and gear box indicates that the proposed method has excellent time- frequency concentration, good noise restraining ability. It can remove noise and reduce misleading interference terms. Therefore, it is more effective for mechanical fault diagnosis.