Abstract: In order to solve the problems of low registration efficiency and poor real-time performance during video stitching, we propose a fast video image stitching algorithm based on grid motion statistics. The algorithm combined the improved Oriented FAST and Rotated BRIEF (ORB) and Grid-based Motion Statistics (GMS) algorithm, and optimized the pyramid scale spatial structure of the image, introduced the fast retinal keypoint algorithm to describe the binary feature points, and used the optimal suture line and Laplacian fusion algorithm to stitch the multi-channel video. Experimental results show that the algorithm improves the robustness and real-time performance of video stitching. Compared with the traditional algorithm, this algorithm improves the image registration speed by 28.7%, 35.2% and 33.9% in brightness, rotation and scale, and improves the accuracy by 21.6%, 10.6% and 9.9%, respectively.