An effective video stitching method

This paper proposes a novel method to create wide-angle and high-resolution videos from video sequences by assembling the individual overlapping video frames. As we know, we register the frame sequences of different videos on the same plane using the perspective matrix and synthesize the overlapped region using a nonlinear blending method which is a time-consuming task; moreover, the scene depth is frequently varying for dynamic video content. Therefore, the projection transform should be changed as the dynamic content in real time. To tackle these problems, we propose a robust algorithm to estimate projection transform fast and blend in real time. Our algorithm includes two stages: the first stage is registration, an accuracy projection transform will get from the first frames of all videos; in the second stage, an optimized method will be used to make the projection transform changed as the sense depth verified in real-time, then a less-exhausted blending method is used to blend images and be gotten a high quality videos. Some real video stitching experiments are carried out, and our experiment results demonstrate that our algorithm can speed up video stitching and get high visual quality obviously.