<p>Electromagnetic tomography (EMT) is a promising tomographic imaging modality that can be used to reconstruct the internal cross-sectional distribution of conductivity and/or permeability of the medium. Due to the ill-posedness of the EMT inverse problem, the reconstructed image is with low quality, even with obvious artifacts and deformations. To further improve the reconstruction quality, we formulate an image reconstruction model for EMT based on a composite regularization framework, which leverages <i>L</i>₁-norm as the residual term to reduce the effect of outliers on the reconstruction, uses <i>L</i>₁ regularization and total variation (TV) regularization as penalty terms to reinforce the sparsity property and preserve edges of the reconstruction objects. Based on the fast iterative shrinkage-thresholding algorithm (FISTA) and the soft thresholding operators, the split Bregman (SB) iteration method is deployed to cope with the multiple regularizers objective function, which splits the complicated imaging inverse problem into several simple sub-problems. Furthermore, an accelerated method is employed to improve the reconstruction speed. Simulation results validate that the reconstruction performance of the proposed method is superior to the classical imaging algorithms, in accordance with the typical phantom objects. The experimental system is constructed to further prove the validity and imaging capability of the presented algorithm, which indicates that the proposed method has potential for promoting the development of EMT imaging and application.</p>