Non-invasive, non-contact, and radiation-free body surface Magnetocardiography (MCG) is capable of accurately capturing the characteristics of the cardiac magnetic field without signal attenuation. It exhibits greater sensitivity to temporal and spatial variances in cardiac electromagnetic behavior compared to conventional ECG signals. Thus, research into MCG Imaging (MCGI) holds the theoretical potential to reconstruct the epicardial electromagnetic activity with higher spatial resolution. No previous studies treated the heart as an epicardial source in MCGI. This study aims to investigate the reliability of MCGI in mapping atrial fibrillation (AF) drivers. In this study, to solve the inverse problem in MCGI, first, P waves or f waves are isolated from MCG signals. The atrium-probe conduction field matrix is then generated using a boundary element algorithm. The ill-conditioned inverse problem is tackled through a multi-scale time-frequency domain regularization scheme. Finally, post-processing of epicardial electromagnetic activity generates the dominant frequency (DF) and entropy maps. Results reveal abnormalities in left and right pulmonary veins (PVs) visible in DF and entropy maps. This finding aligns with the clinical hypothesis suggesting that the PVs serve as high-risk locations for the AF drivers. Furthermore, this study suggests initial potential for MCGI using epicardial sources in AF driver mapping.