A new method for the fault diagnosis of early rub-impact of a rotor by using morphological component analysis(MCA) is proposed in this paper. The proposed method uses MCA to extract impulses from vibration signals of a rotor with early rub-impact fault. According to the morphological difference of each component in the signal, different sparse dictionaries are built by MCA to separate each component from the signal. When the early rub-impact occuring in a rotor system, the vibration signal is often composed of periodic signal with rotor frequency component and its harmonics,transient impulse signal with rotor fault information and random noise. Periodic signals represent the smooth part of the signal, while impulse signals represent the detail part of the signal, therefore, the two kinds of signals can be separated according its morphological difference. Though the improvement of threshold method of MCA, the MCA based on semisoft shrinkage is proposed. The simulation result shows that the MCA based on semisoft shrinkage is superior to the MCA based on hard threshold. Analysis result of the vibration signal of early rub-impact of a rotor shows that the MCA based on semisoft shrinkage is effective in extracting impulse component from the vibration signal of early rub-impact of a rotor, and diagnosis the fault of early rub-impact of a rotor.