Scalable Parallel Geometric Hashing for Hypercube SIMD Architectures

摘要

Geometric hashing has recently been introduced as a new paradigm for model based object recognition. The geometric hashing algorithm allows one to find instances of model point patterns in a scene, subject to noise, obscuration, and transformation. In this paper we concentrate on the cases of similarity and rigid transformations and examine the parallelizability of the algorithm. We describe two scalable algorithms for hypercube SIMD architectures. A number of important building block algorithms and several variations to the basic approach are discussed.