When rolling element bearing occurred incipient fault,its vibration signal from bearing seat is usually very weak,and the hidden impulse components may be masked by machine noise or background noise,making it difficult to extract the fault features by frequency spectrum or envelope spectrum analysis. Through the method of minimum entropy deconvolution（ MED）,the signal to noise ratio can be improved,and the impulse components from fault signal can be highlighted. The MED method is applied to corresponding deconvolution filtering for three typical early fault signals from rolling element,outer race,and inner race. Then,through the amplitude spectrum and envelope demodulation analysis,the corresponding typical fault features of rolling element bearing are successfully extracted. The effectiveness and advantage of MED method in incipient fault diagnosis of rolling element bearing is verified by the application results.