“Light Waving: Estimating Light Positions From Photographs Alone”

We present an algorithm to automatically estimate three- dimensional light positions from an unordered set of images. Image-based rendering and relighting are used to create highly realistic images with arbitrary illumination without the need to compute light-transport from first principles. Most of these techniques operate on a linear combination of input images, taken under varying lighting conditions, for which the light positions have to be known. Establishing these positions generally involves large, complicated equipment or extensive calibration procedures. Instead, we apply dimensionality reduction directly to the input im- ages to estimate light positions. This obviates the need for calibration and specialized equipment, making our approach inexpensive, portable and applicable to objects of almost any size.

Tenenbaum, J. B., de Silva, V., and Langford, J. C. 2000. A global framework for nonlinear dimensionality reduction. Science 290, 5500 (December), 2319–2323.