Drift-free Motion Estimation from Video Images using Phase Correlation and Linear Optimization

Motion reconstruction from video images is known as an effective way in the autonomous robot guidance or teaching for a particular task. The authors' goal is to reconstruct motion from a video with the image-based visual servoing, and the main topic of this paper is to estimate camera motion on the image plane to make a reference. This paper focuses on a drift-free robust camera motion estimation by utilizing all relationships between every two frames. The proposed method in this paper does not use feature points like well-known bundle adjustment. It enables us to use a linear equation to solve its optimization. Then, brand new solution by using distance matrix is proposed so that it can save computation time and memory against a common pseudo inverse matrix based method. Finally, both qualitative and quantitative evaluations based on image mosaicing technique were achieved to confirm the effectiveness of proposed method.