Fast construction of dynamic and multi-resolution 360° panoramas from video sequences

This paper presents a new approach to automatically build a dynamic and multi- resolution 360( panoramic (DMP) representation from image sequences taken by a rotational and zoom camera. A cylindrical panorama is generated by mosaicing the image sequence taken by a hand-held camera. A multi-resolution representation is built for the more interesting areas by means of camera zooming. The dynamic objects in the scene can be detected and represented separately. Although a simple (yet stable) rigid motion model is used to estimate inter-frame motion parameters, the mosaicing methodology presented in this paper enables precise mosaicing. Moving objects are detected and separated from images based on motion information, and more accurate contours are extracted using a modified active contour algorithm. The DMP construction method is fast, robust and automatic, achieving 1 frame per second in a 266MHz PC. No camera calibration, feature extraction, image segmentation or complicated nonlinear iterative processing is required in our algorithms. The construction of the DMP representation can be used in virtual reality, video surveillance and very low bit-rate video coding.