A Fast Video Mosaic Algorithm Based on Optimal Homography Matrix
Wen Wang, Chunyu Zhang, Li Sang, Yong Jun Su, Ke-you Guo · 2012
Mosaic technology is widely used in traffic surveillance. In this paper, we propose a fast video mosaic algorithm based on optimal homography matrix. First, we extract characteristic feature points using SURF method. Second, we match corresponding feature points using NN (Nearest Neighbor) and RANSAC (Random Sample Consensus) method. Third, we calculate the optimal homography matrix using the normalized covariance correlation of the overlapped area between corresponding video frames. Lastly, we obtain the mosaic video image using the optimal homography matrix. Experiments in traffic video surveillance show that our algorithm can meet the needs of real-time requirements with processing speed per frame about 40 ms in average. It is more robust, has higher accuracy, has lower computational complexity, and has faster operational speed. So, it has value in both theoretical research and practical application.