Fast Video Stabilization Algorithms

A set of fast and robust electronic video stabilization algorithms are presented in this thesis. The first algorithm is based on a two-dimensional feature-based motion estimation technique. The method tracks a small set of features and estimates the movement of the camera between consecutive frames. An affine motion model is utilized to determine the parameters of translation and rotation between images. The determined affine transformation is then exploited to compensate for the abrupt temporal discontinuities of input image sequences. Also, a Frequency domain approach is developed to estimate translations between two consecutive frames in a video sequence. Finally, a jitter detection technique has been developed to isolate vibration affected subsequences from an image sequence. The experimental results of using both simulated and real images have revealed the applicability of the proposed techniques. In particular, the emphasis has been to develop real time implementable algorithms, suitable for unmanned vehicles with severe payload constraints.