In this study, in order to eliminate the stripping error from GRACE gravity observations and localization of the signals of this satellite, wavelet analysis instead of Gaussian isotropic filtering approach is used. For this purpose, the scaling functions and wavelets, generated by a so-called cubic polynomial (CuP), are considered and gravity variations for two major earthquakes-Sumatra earthquake and Maule earthquake- are obtained. For the case of Sumatra earthquake- Indonesia , various scale function of CuP are tested and the most precise results of gravity variations with maximum 30 Micro-Gal is acquired in scale 4 which was similar to other studies. This result  didn’t change with increasing scale function even higher than 4. Moreover, for the case of Maule earthquake- Chile, the gravity variations before and after of using wavelet is attained which shows many disparity. The maximum gravity variations before applying wavelet is acquired 20 Micro-Gal and after applying that, the maximum of these variations is computed 10 Micro-Gal which was similar to other researches.