Research on background motion estimation and compensation in image sequences

In the dynamic image sequence, there exists the camera´s movement when we are capturing, so when we analyze the image sequences, we must do the global motion estimation and compensation at first. Adopting the feature matching method, we make use of the feature points to estimate the motion vector in this article. These points are picked up by the improved Minimum Intensity Change algorithm, and use the Adaptive Rood Pattern Search to find out the matching points of each feature point, then combine with the Random Sample Consensus algorithm and least square method to estimate the affine transformation parameters of the background motion model, use the bilinear interpolation to realize the technology of the background motion compensation at last. Finally, we do the experiments to verify the method used in this paper and analyze its feasibility.