Fake Fur Rendering

Dan B Goldman, "Fake Fur Rendering", In SIGGRAPH 97 Conference Proceedings, pp. 127-134., 1997.


A probabilistic lighting model is presented for thin coats of fur over skin. Previous methods for rendering furry objects and creatures have addressed the case where individual strands or tufts of hair may be resolvable at the pixel level. These methods are often compuationally intensive. However, a large class of real-world cases where individual hairs are much smaller than the size of a pixel can be addressed using a probabilistic model for the expected value of reflected light within a small surface area. Under the assumption that hair parameters are slowly varying across the skin, lighting calculations are performed on a reference hair with prefiltered parameters. The reflected light from individual hairs and from the skin below is blended using the expectation of a ray striking a hair in that area as the opacity of hte fur coating. Approximations for hair-to-hair shadowing and hair-to-skin shadowing can be made using the same hit-expectation model. Our system can be implemented in existing commercial surface-rendering software at a much lower computation cost than typical resolvable-hair methods.


Figures 5-8 were all generated from "film resolution" images. Figures 5 and 6 are the property of Disney and are not redistributable in any form. And, due to confidentiality concerns about proprietary film-recording technology at the time of publication, it has not been possible to release the original image data for figures 7 and 8 at their original resolution. The TIFF files available below for download for those figures were rescanned from photographic prints on a consumer-grade flatbed scanner (circa 1997!).