In this paper we present a robust technique for summarizing sport video sequences. Unlike dense optical flow and parametric methods, we develop a semi-automatic application that uses geometrical information, such as straight lines, extracted from sequences of images. This information is used to compute the homographies between consecutive frames. This estimation yields a manner of synthesizing a high resolution image of the background plus the field-centered trajectories of the moving objects onto it. © Springer-Verlag Berlin Heidelberg 2003.
|Journal||Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)|
|Publication status||Published - 1 Dec 2003|