Pixar's REYES rendered complex scenes by shrinking everything to micropolygons

Pixar's 1987 REYES paper attacked a rendering problem that brute-force geometry pipelines handled badly: complex film scenes with many surface types, textures, and visibility cases.

The clever move was to stop treating each object as special. REYES reduced all objects to micropolygons, then let shading and visibility operate on that shared unit in the coordinate system best suited to each calculation.

That architecture made arbitrarily complex models manageable because locality minimized paging, texture filtering became simpler, and screen-space stochastic sampling handled visibility without awkward clipping and inverse perspective work.

Reduce every shape to one tiny common unit

A shared micropolygon representation moved complexity out of the later rendering stages.

Different coordinate systems were used where they made the math easier, instead of forcing one global space.

The architecture exploited locality, so huge scenes could be processed in small pieces.

When inputs are wildly different, find the common intermediate form that makes every later operation simple.

The paper became a core computer-graphics reference, with OpenAlex listing it as a 1987 ACM SIGGRAPH Computer Graphics article by Pixar authors and recording 237 citations.