An accurate orientation estimation is a primary requirement in autonomous navigation tasks of mobile robots, especially when such operation is involved in the solution of the position estimation problem (e.g. dead reckoning navigation systems). This is the case of Autonomous Underwater Vehicles (AUVs), subject of this paper; for this class of robots, a complete attitude estimation system based on the adaptation of a Nonlinear Complementary Filter (NCF) is proposed. The presented filter fuses the measurements of an Inertial Measurement Unit (IMU) comprising magnetometers. The integration of a single-axis Fibre Optic Gyroscope (FOG) is evaluated to overcome issues due to magnetic disturbances. In addition, a calibration procedure for magnetometers to be exploited within the attitude estimation filter is proposed. The whole system has been experimentally validated and tested on the University of Florence Typhoon AUV; the results prove that the proposed approach allows to obtain a reliable orientation estimate even in the presence of external unpredictable disturbances.