A research team from Chiba University, Kyoto University, and Nara Institute of Science and Technology has successfully developed a new mathematical model for computer graphics (CG) that can calculate multiple reflections with extremely high accuracy while significantly reducing image noise (variation). This model can generate high-quality images in a short computation time in all fields where realistic CG images are required, such as movies, games, product design, and architectural visualization.
To realistically depict materials such as metal and plastic in computer graphics, it is necessary to physically and accurately calculate how light reflects off the microscopic irregularities (microsurfaces) on the surface. Several methods have been proposed to calculate "multiple reflections," where light reflects "more than once," in order to physically and accurately reproduce the texture of an object's surface. However, these methods often result in a lot of noise in the image, or at least significant errors (biases) in the calculation results.
Therefore, the research group formulated the optical path while partially retaining the "height" information of each reflection point (height-dependent optical path formulation), and then introduced a method called "sequential height pre-integration" which involves stepwise integration in the height direction, thereby converting it into a form that can be efficiently computed. By combining this with "path type-specific BRDF evaluation," which assigns the optimal number of computational samples for each type of optical path, they were able to suppress image noise.
The experimental results confirmed that, compared to existing representative methods, this method can reproduce multiple reflections on rough metal surfaces with less noise and extremely small errors in the same computation time. It is expected that this will contribute to accelerating and reducing the cost of producing high-quality CG images used in movies, games, and product design.
