We consider a regularized Maximum Likelihood Estimation (MLE) framework to produce images in the context of radio interferometric measurements. Specifically, we consider the class of compound Gaussian distributions to model the additive noise in the presence of radiofrequency interferences. In most cases, direct maximization of the likelihood is not tractable. To overcome this issue, we propose a generic expectation-maximization (EM) algorithm in the presence of a compound Gaussian noise. In addition, we leverage an approximation of the forward radio interferometric operator to derive an original latent data space that allows the use of the FFT in the maximization step, leading to an accelerated extension of the proposed imaging algorithm. The proposed approaches are evaluated on simulated and real data and show a significant improvement in the robustness to the presence of radiofrequency interferences (RFI) in the measurement.