Realistic rendering 3D IFS fractals in real-time with graphics accelerators

In this paper we present a novel approach to realistic real-time rendering scenes consisting of many affine IFS fractals. In order to illuminate the fractals, we propose a new method for estimating normals at fractal surface points. The method is based on approximations of the convex hulls for fractal subsets. The geometry of a fractal is represented as a collection of circular splats. We also show how to take advantage of self-similarity and hardware geometry instancing so as to store fractal models with extremely small memory requirements. As a result, the geometry data related to hundreds of fractals can be kept entirely in the video RAM of a graphics adapter and rendered in real-time using even medium-power graphics accelerators.