This paper proposes a method to compute the DWT (discrete wavelet transform) coefficients using Mallat's fast algorithm with Daubechies N (N=1,…,10) wavelet series as the wavelet primary function. The authors verified the reliability of the method with a theoretical model. By using an approach to multi-layer decomposition and a strategy of threshold value in the process of the airborne gravity data denoising, the authors extracted the free air gravity anomalies by the frequency and bandwidth character of the gravity anomalies. A comparison is made by using six, seven and eight layers decomposition and by employing the soft-threshold and the hard-threshold denoising. The results demonstrate that the free air gravity anomalies extracted by this method are quite consistent with the filtered results of the GT-1A system.