Real-Time Image Stabilization for Digital Video Cameras

Real-time image stabilization is particular useful for portable video cameras to compensate the vibrations caused by the photographer. Although several stabilization algorithms have been proposed, very few of them consider the practical implementation issues in a camera platform but only simulate the algorithms in computers. In this paper, we propose a robust real-time algorithm to estimate global motion vector (GMV) and stabilize the video frames by fully utilizing the available computational engine in a typical camera system-on-a-chip (SoC) platform. The resulting video is much stable than the one without applying stabilization, and the speed is fast enough for real-time video recording applications