Perceptual illumination components: A new approach to efficient, high-quality global illumination rendering.

Master's thesis, Cornell University, August 2004.

We introduce a new perceptual metric for efficient, high quality, global illumination rendering. The metric is based on a rendering-by-components framework in which the direct, and indirect diffuse, glossy, and specular light transport paths are separately computed and then composited to produce a high quality image. The metric predicts the perceptual importances of the computationally expensive indirect illumination components with respect to image quality. To develop the metric we conducted a series of psychophysical experiments in which we measured and modeled the perceptual importances of the components. An important property of this new metric is that it predicts component importances from inexpensive estimates of the reflectance properties of a scene, and therefore adds negligible overhead to the rendering process. This perceptual metric should enable the development of an important new class of efficient globalillumination rendering systems that can intelligently allocate limited computational resources, to provide high quality images at interactive rates.

This paper is available as a PDF file Sto04.pdf ( 13M).