Mesh Simplification and Decimation
Reducing a mesh's triangle count while preserving its shape keeps 3D models interactive and enables level-of-detail rendering.
Fewer triangles, same shape
Detailed geometry and extracted isosurfaces often contain far more triangles than a screen can use or a GPU can render fast. Mesh simplification (decimation) removes triangles while keeping the visible shape, trading imperceptible detail for speed. It is the geometric side of level-of-detail.
How it works
The dominant method is iterative edge collapse: repeatedly merge the edge whose removal changes the surface least, measured by an error metric. The quadric error metric, which sums squared distances to the planes of adjacent faces, is the standard because it is fast and preserves sharp features and silhouettes well.
- Collapse low-error edges first to protect important detail.
- Preserve boundaries, sharp edges, and features that carry meaning.
- Generate a hierarchy of levels for distance-based swapping.
Error control
Simplification must not lie about geometry. Bounding the error keeps the simplified mesh within a stated tolerance of the original, so decisions made from it remain valid. For engineering models, some features (a critical clearance, a mating surface) may be marked to never simplify.
LOD hierarchies
Producing several decimated levels lets a renderer show a coarse mesh at distance and a fine one up close (see Level of Detail), holding frame rate steady as the camera moves.
Kronos use
Public browser models are decimated to interactive sizes with bounded error, and the public build strips detail that is not to be released while keeping overall geometry faithful.