Global motion estimation based on SIFT feature match for digital image stabilization

This paper proposes a global motion estimation method based on SIFT feature match according to the characteristics and requirements of digital image stabilization. This method proceeds by using multi-block when extracting key points, followed by in-block match selection using reverse match, and finally inter-block match selection using angle invariance of similarity transformation. The experimental results show that this method with low complexity and high accuracy can achieve good performance dealing with variety of videos, and is proved useful in digital image stabilization technology.