Imaging beyond the Pinhole Camera

The world’s first photograph was taken by Joseph Nicphore Niépce (1775–1833) in 1826 on his country estate near Chalon-sur-Sane, France. The photo shows parts of farm buildings and some sky. Exposure time was eight hours. Niépce used a pinhole camera, known as camera obscura, and utilized pewter plates as the support medium for the photographic process. The camera obscura, the basic projection model of pinhole cameras, was first reported by the Chinese philosopher Mo-Ti (5th century BC): light rays passing through a pinhole into a darkened room create an upside-down image of the outside world. Cameras used since Niépce are basically following the pinhole camera principle. The quality of projected images improved due to progress in optical lenses and silver-based film, the latter one replaced today by digital technologies. Pinhole-type cameras are still the dominating brands, and also used in computer vision for understanding 3D scenes based on captured images or videos. However, different applications have pushed for designing alternative architectures of cameras. For example, in photogrammetry cameras are installed in planes or satellites, and a continues stream of image data can also be created by capturing images just line by line, one line at a time. As a second example, robots require to understand a scenery in full 360◦ to be able to react on obstacles or events; a camera looking upward into a parabolic or hyperbolic mirror allows this type of omnidirectional viewing. The development of alternative camera architectures also requires to understand related projective geometries for the purpose of camera calibration, binocular stereo, or static or dynamic scene understanding. This abstract collection reports about contributions given at a workshop at the international computer science center in Dagstuhl (Germany) addressing basics and applications of alternative camera technologies, in particular in the context of computer vision, computer graphics, visualisation centers, camera producers, or application areas such remote sensing, surveillance, ambient intelligence, satellite or super-high resolution imaging. Examples of subjects are geometry and image processing on plenoptic modalities, multiperspective image acquisition, panoramic imaging, plenoptic sampling and editing, new camera technologies and related theoretical issues. This abstract collection is structured into five parts on (1) sensor geometry for different camera architectures, also adressing calibration, (2) applications of non-pinhole cameras for analyzing motion, (3) mapping of 3D scenes into 3D models, (4) navigation of robots using new camera technologies, and (5) on specialized aspects of new sensors and other modalities. New results and specific research strategies have been discussed at this seminar to approach this highly complex field. The seminar intention was to discuss theoretical fundamentals related to those issues and to specify open problems and major directions of further development in the field of new camera technologies related to computer vision, computer graphics and related applications. The seminar schedule was characterised by flexibility, working groups, and sufficient time for focused discussions. The participants of this seminar enjoyed the atmosphere and the services at Dagstuhl very much. The quality of this center is unique. There will be an edited volume of seminar papers (within the Kluwer series) with an expected publication date in early 2006.