Subsurface Texture Mapping

Feature Article C areful modeling and rendering of natural materials is essential for creating realistic images. Graphic designers often model the interactions between light and materials using a bidi-rectional reflectance distribution function (BRDF), which assumes that a light ray enters and exits the material's surface at the same point. This assumption doesn't hold for materials such as marble, wax, milk, leaves, and human skin. For these translucent materials , light enters the surface and scatters inside the medium before leaving the surface. Because of this light interaction, translucent objects have a smooth and soft appearance. In addition, subsurface scattering produces interesting color shifts. In human skin, scattering is primarily limited to the red wave length, which is why we see a reddish glow. In complex organic materials, this phenomenon is also responsible for exhibiting blurry details of the inner composition, such as for veins under the human skin. Because such materials are common, their accurate modeling and rendering is important. This article presents a rendering technique for multilayered materials. Unlike existing methods, our technique doesn't assume that the thickness of these layers is constant (see the " Related Work " sidebar). We use relief texture mapping 1 to model a material's interior. Instead of representing the surface details, we use this method to represent the object's inner structure. We describe the material's layers using a simple 2D texture, in which each channel encodes a thickness. Our method supports nonplanar surfaces and falls between subsurface rendering methods based on surfaces and 3D texture-based algorithms. Furthermore, our solution provides a compact way to design translucent objects using a small amount of data. Overview Unlike methods that use planar layers, ours computes single scattering into layers of variable thickness (see Figures 1a and 1b). To provide a real-time but realistic rendering, we limit our computation to single scattering. Our method, implemented on graphics hardware, uses a ray-marching algorithm, illustrated in Figure 1c. To model layered material, we need information about the layers' thickness, which varies beneath the surface. Computing subsurface scattering within this kind of material requires knowing the layers' position and scattering properties. To this end, we propose a point-localization method. As Figure 1c shows, we approximate the reflected luminance at point P due to single-scattering events occurring along the viewing ray underneath the object's surface. We compute the contributions of a certain number of sample points M on the viewing ray …