Video Mosaic System Based On Feature Points

With the development of computer graphics, image stitching technology have provide an effective solution for the realization of high-resolution images with a wide viewing angle. Image stitching technology will help a series of images which have some overlapping regions to generate a high-resolution image with wide angle by certain means. Video stitching technology is developed by image stitching technology, which taking the real-time effective into account. Video stitching has become the key technology in the field of computer vision, image processing, and virtual reality, also been widely used in space exploration, medical image analysis, real-time television and road monitoring and other fields.The process of video stitching techniques are:input images, geometric correction, image matching, image fusion then generate the mosaic video.In order to achieve the goal of video splicing technology, this paper conducted a series researches on these issues. In the stage of image matching, the classical features detection algorithm such as: Moravec corner points detection algorithm, Harris corner points detection algorithm, Forestener corner points detection algorithm, SIFT feature points detection algorithm and SURF feature points detection algorithm have been studied. By the comparison of simulation results of different algorithms to decide which feature point detection algorithm will be selected in the video mosaic system.In the image fusion stage, using the optimal stitching line to fusion image, through the light may have a significant transformation, stitching image still has a good visual effect.On this basis, in order to solve the possible move between cameras during the video capture time, this paper proposed a method for dynamic video stitching to avoid errors. Use color information of image to decide whether to re-calculate the transform matrix to achieve real- time video stitching system.Finally in the real system have realized the above studies, result proves that this system has a good problem solving results for video mosaic process,through there may exist changes the relative position between cameras, still this system not only has the noise robustness, but also has scale invariance, illumination invariant features.