| Now the video capture device is very common, it has been entered every household already. With the development and progress of social, The need of the amount of information increased quickly. It is also means the need of breadth and continuity is growing fast, too. However, the image, which collected by a single camera, has limitations in range and angle. So a completely and continuously image which synthesized by a plurality of video images was very desired in fact. Now the splicing technology of static image is relatively mature, but the research and application of video splicing technology are not very well.In this paper, a real-time algorithm based on sift algorithm has been put forward, when two camera were fixed. The algorithm can be divided into two stages. The first is to calculate the transformational matrix. And the second is to realize the real-time splicing. The consuming time of video stitching was from lift the feature points of the first frame to the successful of match and stitching, So the video stitching divide into two stages. The sift algorithm has been used to lift the feature points from the first image. Using the projection transformation model and Ransac method, the transformation matrix of the first couple frames can be calculated, when purified feature points. Through experimental analysis, under the condition of two cameras were fixed, the transformation matrix of subsequent frames is similar to the transformation matrix of the first couple, and error was very small.The result of using the transformation matrix of first couple frames to stitch subsequent frames of videos is good, so the transformation matrix of first couple can instead of the others. This will prevent the repeated extraction and matching of feature points and time redundancy of computing related transformation matrix. The experiments proved that the splicing effect is good, the video stitching of real-time processing can be achieved with the0.0387/frame stitching speed... |
PDF Full Text Request