Raymarching Distance Fields

最近在研究体积云，看到了一篇介绍raymarching的很好的文章，用它来理解shader toy上的例子也是很好的。特此转载给国内的友人们。

顺便再附一张用这种技术做出的体积云效果，还讲究吧 。继续研究ing

Raymarching Distance Fields

Hello. This is my first article in a series of many, where I dive deep into an interesting algorithm and investigate what it can be used for, and (hopefully) end up with something nice to show in the end. For each of the projects my target result is a technical writeup like this, and either an interactive demo or some cool visuals. All code will be uploaded to Github, with full documentation.

For this article I will write about the rendering technique known as raymarching with distance fields, capable of producing highly detailed images in real-time with very simple code.

slisesix by Inigo Quilez

Before reading on, perhaps you would like to try the interactive WebGL demo?

Content

• Introduction
• The raymarching algorithm
• Rendering techniques
  • Iteration based coloring
  • Bounding volume
  • Fog
  • Anti-aliasing
  • Lighting
  • Shadows
  • Ambient occlusion
  • Reflections
  • Distortion and repetition
• Other
• Conclusion
• Code
• References

Introduction

Raymarching is a 3d-rendering technique, praised by programming-enthusiasts for both its simplicity and speed. It has been used extensively in the demoscene, producing low-size executables and amazing visuals. The most frontstanding figure behind its popularity, is Inigo Quilez, promoting it with his presentation at nvscene:Rendering Worlds With Two Triangles.

The idea is this: Say you have some surface in space. You don't have an explicit formula for it, nor a set of triangles describing it. But you can find out how far away it is, from any point. How would you render this surface?

First of all, we need to find out which points that lie on the surface, and what pixels they correspond to. To do this we use a technique known asray-casting.

Imagine you and your monitor being placed in this virtual world. Your eye will be looking at a rectangle (your monitor), which we shall call the image plane. Ray-casting works by shooting rays from the eye through each pixel on the image plane, and finding the closest object blocking the path of the ray. Once we hit an object, we can compute the color and shading of the corresponding pixel. If the ray does not hit anything, the pixel is colored with somesky color.

There are several ways in which we can calculate the intersection, for example we analytically solve for it. A raymarcher, however, looks for an approximate solution, by marching along the ray in steps until it finds an intersection. By controlling the step size using a distance field, we can reach blazing speeds, even on a regular laptop GPU.

The raymarching algorithm

In traditional raytracing, a scene is often described by a set of triangles or spheres, making up a mesh. Using some spatial acceleration structure, we can quickly solve for the exact intersections between the rays and the objects.

With raymarching however, we allow for some leeway in the intersection, and accept it when a ray isclose enough to a surface. This is done by marching along the ray at step sizes, and checking whethe