Video Image Stitching Technology Based On SIFT Algorithm

Video stitching has great significance in real life. It due to 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 collects by a single camera has the limitation in range and angle. So a completely and continuously image which synthesized by a plurality of video images was much desired in fact. Now the splicing technology of image is relatively mature, but the research and application of video splicing technology are not very well. So this is the significance of the method in this paper.In this paper, a real-time algorithm based on sift(Scale Invariant Feature Transform) algorithm has been put forward; it mainly uses the transform matrix. Feature points matching algorithm based on SIFT algorithm with strong robustness has better match capability.The algorithm can be divided into two stages. The first part is to realize the stitching of a couple of images and calculate the transformational matrix. The second part is to use the transformation matrix to realize the stitching of all frames of video. So we can get the seamless stitching video. Among them, the first part is aimed at the stitching of image and optimizes the algorithm. This method first utilizes SIFT algorithm to extract feature points and applies angle cosine to do coarse matching between feature points, then uses RANSAC algorithm to eliminate false matching points. Using the projection transformation model and Ransac method, the transformation matrix of the first frame can be calculated. This video stitching method applies image registration and fusion for the first frame rather than each frame. It takes too much time. So in the second part, the key is to get the transformation matrix of the first frame, and then apply the transform matrix to fulfill the stitching of the video. So that it can increase the speed of video stitching. At last the image fusion utilizing gradually fading out algorithm is also taken into account. Considering the image brightness differences caused by the light when capturing the video, light compensation is adopted to process images at the beginning. The experimental results show that this method can realize the stitching of the real-time and seamlessly clear wide scene video.