Aiming at the problem that the incipient fault feature of rolling bearing is easily submerged in the environmental noise and is difficult to extract,an incipient fault diagnosis method combined minimum entropy deconvolution（ MED） denoise with variational mode decomposition（ VMD） is proposed. Firstly,the bearing vibration signal is denoised by using MED with kurtosis as maximization criteria. Then,the processed signal is decomposed into several components through VMD which is a new high precision multi-component decomposition method. At last,the bearing fault is diagnosed by analyzing fault frequency in the envelope spectrum of the maximal kurtosis component. The experimental signal analysis results show the validity of the proposed method.