Purchase this article with an account.
Yaniv Morgenstern, Richard F. Murray, Wilson S. Geisler; Real-world illumination measurements with a multidirectional photometer. Journal of Vision 2010;10(7):444. doi: 10.1167/10.7.444.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
The visual system resolves ambiguity by relying on assumptions that reflect environmental regularities. One well-known assumption, used to interpret ambiguous 2D images, is the light-from-above prior. However, recent work has shown the visual system represents more complex assumptions of natural illumination than a single overhead light source (Fleming et al., 2003; Doerschner et al., 2007). To investigate these hidden assumptions previous researchers have used multi-directional photographic methods to measure and statistically characterize natural illumination. These methods provide high-resolution, high-dynamic range images of the complete surrounding scene. For some purposes, such as understanding illumination of Lambertian objects, a coarser lighting measurement that represents the first three orders of spherical harmonics would suffice (Basri and Jacobs, 2001; Ramamoorthi and Hanrahan, 2001). We will describe a multidirectional photometer we have developed, that makes fast and accurate measurements of low-degree spherical harmonic components of real-world lighting. The multidirectional photometer is a 20 cm diameter aluminum sphere, mounted with 64 approximately evenly spaced photodiodes. Each photodiode is filtered to match the photopic spectral sensitivity of the human visual system, and fitted with an aperture that reduces its directional selectivity so as to provide the sharpest image possible with 64 sensors. The device measures light ranging from low-lit indoor scenes to direct sunlight, and makes several complete measurements per second. We discuss design decisions such as how many photodiodes to use, how to distribute the photodiodes over the sphere, and what directional selectivity to give the individual photodiodes. We also discuss a linear-systems approach to using the photodiode measurements to reconstruct ambient lighting as a sum of basis functions. We will present preliminary findings on the statistical characterization of low-degree spherical harmonics of light in natural scenes, and discuss how such measurements can be used to advance our understanding of shape from shading and lightness constancy.
This PDF is available to Subscribers Only