In the ISAR imaging of targets with rotating parts, the image of the targets main body may be shadowed by the micro-Doppler effect induced by the rotating parts. To solve this problem, an algorithm utilizing complex local mean decomposition is researched to separate the micro-Doppler signature. The Doppler frequency difference between the main body and the rotating parts is analyzed, and the complex local mean decomposition method is utilized to separate the micro-Doppler signature which accurately separate the oscillation mode embedded in signals and decompose the complex modulation non-stationary signals adaptively into some stable mono-components. After the separation, better geometrical features of the main body and micro-Doppler features of the rotating parts can be obtained by processing each independently. The results from the simulated and measured data are given to verify the validity of the proposed algorithm.