The Sensor Pattern Noise (SPN) is easily contaminated by CFA interpolation noise and JPEG compression noise. These unwanted artifacts may result in false identifications in SPN-based source camera detection. A preprocessing method of SPN based on spatial smoothing was proposed in order to improve the accuracy of source camera detection. It is assumed that the SPN is a random signal similar to white Gaussian noise (WGN), and has a flat frequency spectrum similar to that of WGN in the frequency domain. Based on this assumption, WGN-guided filtering was used in the spatial domain in the proposed method. The spatial smoothing effect can suppress the unwanted artifacts of the SPN. At the same time, the SPN can maintain its own properties and possess similar characteristics with WGN. The proposed method was evaluated on our own database of camera images from cellphones. The experimental results show that the proposed method outperforms the existing preprocessing methods, the accuracy of the source camera detection being increased by more than 0.026 in terms of the Kappa statistical coefficients. The robustness to JPEG compression of the proposed method is also better than others.