Based on the thought of least square, Full Waveform Inversion (FWI) has the characteristics of high resolution and high precision, because it considers both amplitude and phase information of seismic waves. However, it will cost large computation, if we directly use FWI in seismic data processing. In order to solve this problem, by applying linear time-shift function to compress a large number of shot gathers into several plane-wave gathers, we realized a method of multi-scale FWI in time domain based on plane-wave encoding. Then we used Marmousimodel to test our method. Comparing the results of our method with conventional FWI results, it showed that our method had the advantage of high computation efficiency and the great ability to suppress noise.