H.264 Sensor Aided Video Encoder for UAV BLOS missions C. V. Angelino*, L. Cicala*, M. De Mizio*, P. Leoncini*, E. Baccaglini**, M. Gavelli**, N. Raimondo**, and R. Scopigno** * CIRA, the Italian Aerospace Research Center, Capua, Italy {c.angelino, l.cicala, m.demizio, p.leoncini}@cira.it ** Istituto Superiore Mario Boella, Torino, Italy {baccaglini, gavelli, raimondo, scopigno}@isbm.it Abstract. This paper presents a new low-complexity H.264 encoder, based on x264 implementation, for Unmanned Aerial Vehicles (UAV) ap- plications. The encoder employs a new motion estimation scheme which make use of the global motion information provided by the onboard nav- igation system. The results are relevant in low frame rate video coding, which is a typical scenario in UAV behind line-of-sight (BLOS) missions. 1 Sensor Aided Global Motion Estimation Global motions in a video sequence are caused by camera motion, which can be modeled by parametric transforms. The process of estimating the transform parameters is called global motion estimation (GME) and compensation (GMC). A widely used global motion model is perspective model with 8 parameters. This model is suitable to represent the motion field for far backgrounds. In order to reduce the computational burden of GME, several works on the use of external motion sensors have been recently proposed. In [2], for example, the authors studied the application of GME techniques that process sensor data (gyroscopes, accelerometers and magnetometers) to video coding. The focus is on mobile devices and the complexity reduction of motion estimation reflects into increasing battery life. Video coding systems currently employed in NATO military UAV applica- tions meet the STANAG 4609 standard [5], which refers to MPEG4 and Motion JPEG2000 standards for motion imagery coding. However, these standards are for general purpose applications and do not rely on the particular features of aerial video sequences. Indeed, the latter present some peculiarities in terms of observed scene geometry and point of view. In particular, the observed scene is almost static (the camera movement is dominant) and generally easy to model (usually approximated with a plane). Moreover, the point of view (camera po- sition and orientation) is far from the scene and approximately known since it can be obtained from the navigation system. UAV are used typically for surveillance, for both military and civil appli- cations such as the remote monitoring of hostile area, coasts and boundaries