Computer animators and video game designers may soon have a better way to create the purple-green sheen of a grackle's wing, or the pink flash on a hummingbird's throat, thanks to a new method for rendering iridescent feathers.
Researchers in the lab of Steve Marschner, professor of computer science in the Cornell Ann S. Bowers College of Computing and Information Science, in partnership with technology company NVIDIA, modeled how light reflects off different types of iridescent bird feathers. Using specimens from the Cornell Lab of Ornithology as a reference, they rendered digital images that change colors at different angles and, unlike earlier models, incorporate tiny variations that give real feathers their glittery appearance.
"Natural iridescent things will always have these imperfect qualities," Marschner said.
Yunchen Yu, a doctoral student in the field of computer science, presented the research, "Appearance Modeling of Iridescent Feathers with Diverse Nanostructures," Dec. 6 at SIGGRAPH Asia 2024, where it won the Best Paper Award.
Yu has long been fascinated by biological iridescence - the sparkling sheen on certain fish, frogs and feathers. While most colors come from pigments, iridescence results from light interacting with nanostructures inside cells - which can only be seen under a microscope - that selectively reflect certain wavelengths.
Yu and her collaborators chose birds for their first attempt at modeling biological iridescence because feathers are easy to work with.
"Bird feathers have the advantage that the structures are kind of sitting there, right on the surface, so they're not buried under, say, fish skin … or a lot of other structures to simulate," Marschner said. The nanostructures in feathers are located in the barbules - tiny filaments that are too small to see with the naked eye, which project from the barbs, the horizontal strands that attach to the feather's central shaft.
There was just one problem: Yu had a tremendous fear of birds.
The first time she visited the Lab of Ornithology, she had to send in her co-author, Bruce Walter, a research associate at Cornell, to handle and photograph the preserved skins and crop out their heads.
Using diagrams from the ornithology literature, Yu modeled the shape of the barbules with the nanostructures inside, and used the cropped photos to check her results. The study included seven species, which covered the different types of iridescence in birds and whose nanostructures had been determined previously - mallard, rock dove, bronzewing, peacock, magpie, starling and hummingbird.
Next, Yu simulated how each nanostructure scatters the light waves. Her model also incorporated some randomness, to simulate the tiny variations that occur in nature, which gave the renderings realistic glints.
"Barbules are like snowflakes," Yu said. "You can't find two identical ones."
The final step was to render images of the actual feathers. Previously, co-author Andrea Weidlich of NVIDIA had built a module in a graphics rendering software program to build the feathers, from the shaft to the barbule, and Yu integrated the iridescence model.
The process quickly and accurately yields individual feathers and parts of birds that shine and change colors in the light.
The next step will be to incorporate the model into a more artist-friendly interface, researchers said, so designers can apply it in video games and animation.
"Even though it takes a lot of computation to build these models, using them in the end is pretty efficient," Marschner said. "You could definitely imagine getting these into a real-time implementation where you could use them in a virtual environment, or a video game."
Ultimately, this work could also be expanded to render entire birds and other iridescent animals, though additional modeling will be needed to render the effects of the pigments in the skin of animals like octopus and fish.
Yu pointed to another positive outcome from this work: She is no longer afraid of birds. By her third trip to the Lab of Ornithology, she was finally able to handle the specimens. "I think my phobia of birds from childhood is healed after this project," she said.
Eugene d'Eon, a research scientist at NVIDIA, is a co-author on the study. The work received support from the National Science Foundation and NVIDIA Corporation.
Patricia Waldron is a writer for the Cornell Ann S. Bowers College of Computing and Information Science.
