Digital Image Stabilization for Cameras on Moving Platform

Digital image stabilization is an important technique including three parts: global motion estimation, motion filtering and motion compensation. To ensure the visual quality of the whole video, the accuracy of global motion estimation and smooth motion vectors over long images sequence are emphasized. In this paper, we propose a novel stabilization algorithm based on particle filtering. Firstly, Harris point detection is improved by grid sampling to select evenly distributed points. The points are initially matched by feature window matching and further validated by distance criterion to get global motion matrix. Then particle filter is employed to obtain smooth scanning motion vector from original motion curves. The performance of the proposed algorithm is demonstrated through computer simulated experiments. It is shown experimentally that particle filtering can reduce the error variance compared to Kalman filtering.