Abstract: In order to solve problems of seam and ghosting in real-time video mosaics, a video mosaic algorithm based on improved random sample consenus (RANSAC) algorithm was proposed. Firstly, speed-up robust features algorithm is used to extract feature points and find matching feature points in the overlapping area of binocular camera. Then, the improved RANSAC algorithm is adopted to remove the mismatched point pairs and obtain the optimal homography matrix. Finally, the pixels in overlapping areas are processed by dynamic fusion. The experiment results show that using the proposed algorithm can effectively eliminate the seam and ghosting in the video overlapping area, and also ensure real-time performance of the video mosaic system.