Image-Based Modeling, Rendering, and Lighting in Fiat Lux
This animation sketch presents how image-based model-ing, rendering, and lighting were used to create the ani-mation Fiat Lux from the SIGGRAPH 99 Electronic Thea-ter. The film features a variety of dynamic objects real-istically rendered into real environments, including St. Pe-ter?s Basilica in Rome. The geometry, appearance, and illumination of the environments were acquired through digital photography and augmented with the synthetic ob-jects to create the animation. The film builds on the tech-niques of The Campanile Movie and Rendering with Natu-ral Light from SIGGRAPH 97 and 98.
The Imagery and Story
Fiat Lux draws its imagery from the life of Galileo Galilei (1564-1642) and his conflict with the church. When he was twenty, Galileo discovered the principle of the pendulum by observing a swinging chandelier while attending mass. This useful timing device quickly set into motion a series of other important scientific discoveries. As the first to ob-serve the sky with a telescope, Galileo made a number of discoveries supporting the Copernican theory of the solar system. As this conflicted with church doctrine, an elderly Galielo was summoned to Rome where he was tried, con-victed, forced to recant, and sentenced to house arrest for life. Though honorably buried in Florence, Galileo was not formally exonerated by the church until 1992. Fiat Lux presents an abstract interpretation of this story using arti-facts and environments from science and religion.
The objects in Fiat Lux are synthetic, but the environments and the lighting are real. The renderings are a computed simulation of what the scenes would actually look like with the synthetic objects added to the real environments. The techniques we used represent an alternative to traditional compositing methods, in which the lighting on the objects is specified manually.
The environments were acquired in Florence and Rome; the images in St. Peter?s were taken within the span of an hour in accordance with our permissions. To record the full range of illumination, we used high dynamic range photography, in which a series of exposures with varying shutter speeds is combined into a single linear-response radiance image. Several scenes exhibited a dynamic range of over 100,000:1.
The appearance and illumination of each environment was recorded with a set of panoramic images and light probe measurements. Each light probe measurement was made by taking one or two telephoto radiance images of a 2-inch mirrored ball placed on a tripod; each provided an omnidirectional illumination measurement at a particular point in space. Several radiance images were retouched using a special high dynamic range editing procedure and specially processed to diminish glare.
We constructed a basic 3D model of each environment using the Fašade photogrammetric modeling system. The models allowed us to create virtual 3D camera moves using projective texture-mapping, as well as to fix the ori-gin of the captured illumination. The light probe images were used to create light sources of the correct intensity and location, thus replicating the illumination for each envi-ronment. The illumination was used to ?un-light? the ground in each scene, allowing the synthetic objects to cast shadows and appear in reflections. The dynamic objects were animated either procedurally or by using the dynamic simulator in Maya 1.0. Renderings were created on a cluster of workstations using Greg Larson?s RADI-ANCE global illumination system to simulate the photo-metric interaction of the objects and the environments. The final look of the film was achieved using a combina-tion of blur, flare, and vignetting filters applied to the high dynamic range renderings.
Contributors: Christine Cheng, H.P. Duiker, Tal Garfinkel, Tim Hawkins, Jenny Huang, and Westley Sarokin. Sup-ported by Interval Research Corporation, the Digital Media Innovation program, and the Berkeley Millennium project. See also: http://www.cs.berkeley.edu/~debevec/FiatLux