Robust Generation of 360° Panoramic Views from Consumer Video Sequences

This paper presents an algorithm for generation of 360° panoramic views from consumer video sequences. The robust approach considers the limited quality and the various distortions that such material usually contains. In a first stage the focal length of the camera is estimated from estimated warping parameters of the planar perspective motion model. Then the images are projected onto a cylinder surface and the amount of rotation of each image is estimated using a gradient descent algorithm. The presented results prove the suitability and accuracy of the algorithm.