Skip to content
uencoder

Geometric Hashing

Geometric hashing encodes a set of geometric feature points into an invariant hash for point pattern matching and object recognition, invariant to translation and rotation.

Purpose

Fast point-set matching in computer vision, invariant under rigid transforms (translation, rotation, scaling).

Mechanism

A pair of points is chosen as a basis; the remaining points are encoded in the basis coordinate system into a hash table (quantized coordinates). At query time, basis pairs are tried on the unknown point set and matched via table lookup.

Use cases

  • Object recognition
  • Point cloud registration
  • Molecular structure matching

Related tools

References