In this paper, we consider a full-duplex multiple-input multiple-output (MIMO) relaying network with the decode-and-forward (DF) protocol. Due to the full-duplex transmissions, the self-interference from the relay transmitter to the relay receiver degrades the system performance. We thus propose an iterative beamforming structure (IBS) to mitigate the self-interference. In this method, the receive beamforming at the relay is optimized to maximize the received signal-to-interference-plus-noise-ratio (Max-SINR), while the transmit beamforming at the relay is optimized to maximize the signal-to-leakage-plusnoise-ratio (Max-SLNR). To further improve the performance, the receive and transmit beamforming matrices are optimized between Max-SINR and Max-SLNR in an iterative manner. Furthermore, in the presence of the residual self-interference, a low-complexity whitening-filter (WF) maximum likelihood (ML) detector is proposed. In this detector, a WF is designed to transform a colored interference-plus-noise to a white noise, while the singular value decomposition is used to convert coupled spatial subchannels to parallel independent ones. From simulations, we find that the proposed IBS performs much better than the existing schemes. Also, the proposed low-complexity detector significantly reduces the complexity of the conventional ML (CML) detector from exponential time (an exponential function of the number of the source transmit antennas) to polynomial one while achieving a slightly better BER performance than the CML due to interference whitening.