Lucas-Kanade Optical Flow Based Camera Motion Estimation Approach

Abstract

This paper presents an alternative of our previously proposed lightweight camera motion estimation method based on descriptor matching. First, based on the FAST feature and Lucas-Kanade Optical flow approach, we obtain tracked feature point pairs of two frames of input image. Secondly, according to the obtained pixel coordinates of the tracking point pairs, the motion of the camera can be estimated by the epipolar geometry method. The experiment results indicate that our proposed work can provide a 25% improvement in time-efficiency relative to our previous work. So that it can have higher suitability and possibility of implementation in low-power computing platforms.