Ray marching is a class of rendering methods for 3D computer graphics where rays are traversed iteratively, effectively dividing each ray into smaller ray segments, sampling some function at each step. For example, in volume ray casting the function would access data points from a 3D scan. In Sphere tracing, the function estimates a distance to step next. Ray marching is also used in physics simulations as an alternative to ray tracing where analytic solutions of the trajectories of light or sound waves are solved. Ray marching for computer graphics often takes advantage of SDFs to determine a maximum safe step-size, while this is less common in physics simulations a similar adaptive step method can be achieved using adaptive Runge-Kutta methods.
The technique dates back to at least the 1980s; the 1989 paper "Hypertexture" by Ken Perlin contains an early example of a ray marching method.
For simple scenes with basic 3D shapes, ray marching does not have many benefits over ray tracing (which finds intersections without marching through the space). Strengths of SDF ray marching are, for example, when morphing shapes, approximating soft shadows, repetition of geometry, and algorithmically defined scenes.
Signed distance functions exist for many primitive 3D shapes. They can be combined using mathematical operations like Modulo operation and Boolean algebra to form more complex surfaces. For instance, taking the modulus of an SDF's input coordinates tiles its volume across all of space, and taking the maximum of two SDFs gives their volumes' surface of intersection. Because SDFs can be defined for many fractals, sphere tracing is often used for 3D fractal rendering.
|
|