Fast digital image stabilization algorithm based on polar transform and circular block matching

A digital image stabilization algorithm based on polar transform and circular block matching is utilized in this paper to stabilize the videos, taken from rolling airborne platform, which have rapid rotation and arbitrary translation changes between adjacent frames. And circular block matching is used to estimate the global motion parameters, the hierarchical search strategy to reduce the amount of computation. The experimental results show that the proposed algorithm can achieve sub-pixel accuracy and the processing speed of 20 frames per second for 360times264 video sequences on a PIV-3.0 GHz CPU.