Video Mosaic with Motion Estimation

Video mosaic is to make a wide or a panoramic image from video sequence by assembling the individual overlapping video frames. In this paper a technique is presented for video mosaic, which employs motion estimation and robust random sample consensus method. Initial point correspondences between each pair of consecutive frames are computed by block matching method based on phase correlation, then the random sample consensus algorithm is used to identify and reject outliers and the alignment parameters between the consecutive frames are calculated to realize the video image alignment. Some video frames are mosaiced to construct a panoramic image to testify the availability of this approach.