A Camera’s View Forms a 3D Volume
A game camera does not see the whole world at once. Its view can be represented as a view frustum: a pyramid-like volume with its tip at the camera, clipped at the near and far distances. Only objects whose bounds overlap that volume are candidates for drawing.
The game can test an object’s simple enclosing shape—often a box or sphere—against the frustum’s boundaries. If the shape lies entirely beyond one of them, the object can be rejected before the renderer spends time preparing its surfaces, textures, and lighting for the image. Testing a rough boundary is much cheaper than processing every detail of a complex model.
“Out of View” Is Not the Same as “Hidden”
Frustum culling checks whether something falls within the camera’s viewing volume; it does not, by itself, determine whether another object blocks it. A building inside the frustum may still be entirely behind a nearer wall. Games can use occlusion culling—additional checks for objects hidden by other geometry—to skip some of those objects too, but that is a separate problem with its own costs.
Nor does skipping an object’s drawing necessarily mean the game stops simulating it. An enemy may still move or make decisions off-screen, depending on the game’s rules. The immediate saving is narrower: avoid work that cannot affect the current view. As the camera turns, the frustum moves with it, so objects that cross back into the volume become candidates for rendering again.
