Environment Illumination in Image-based Rendering King-Wai Fung

In recent years, image-based rendering has emerged itself as a new and promising methodology in the computer graphics community. One research axis focuses on generating the novel views. Another axis concentrates on the study of varying illumination. However, the lighting conditions in current research are very restrictive, like pure directional or point sources. In this proposal, we describe a new method to efficiently render an image-based object in an environment. Our methodology makes use of spherical harmonics to represent the apparent BRDF of each pixel in an image. After transforming the environment illumination into a spherical harmonic coefficient vector, the new image is computed in the frequency domain. The resulting re-rendered image faithfully represents reflections contributed by the whole surrounding environment. Area light source, which is computationally inefficient in geometry-based computer graphics, can also be faithfully accounted for.