In this paper, we propose a method for time-lapse seismic difference analysis by combining the wavelet transform and nonlinear filtering methods. The nonlinear filter methods we used include median filter and morphological operations, which both show good properties for removing noise and extracting objects. Wavelet transform is a good tool for multi-scale analysis. A combination of the two has the potential to obtain better results than using a single method. We tested our method using both synthetic and real time-lapse seismic data. The results show that the processed time-lapse seismic differences have a clearer background and therefore are easy to interpret.