Video sensor networks with multiple cameras allow to record a scene from different perspectives at the same time. Often, it is desirable to stitch different camera images together to obtain a unified view. We propose an algorithm that allows to stitch views with different centers of projection together into a single panoramic image. A common trajectory is defined in two source images to be merged. Two convex combinations of a warped and a canonic coordinate are applied so that both source images fit together. Applications are manifold, including video surveillance and tracking of objects with multiple static cameras.
Colloquium
