Accelerometer-Assisted Motion Estimation for Efficient Video Encoding

Motion estimation/compensation has been employed in almost all modern video coding standards to effectively reduce temporal redundancy. The encoder complexity is dominated by the motion estimation. In this paper, we present a novel approach to significantly reduce the computation cost of motion estimation in video encoding by employing accelerometers available in many video camcorders and video phones. For proof of concept, we have implemented a working prototype using a camcorder combined with two accelerometers to simultaneously capture video and three-axis acceleration data. The video is then compressed with a reference MPEG-2 encoder, modified to incorporate the accelerometer data to assist motion estimation. We employ a two-stage motion estimation method, consisting of global motion estimation via accelerometers for predicting camera motion and local motion estimation for macroblock prediction. Our experimental data shows a 2 to 3 times speed improvement for the entire encoding process, in comparison with full search.