Applying graphics hardware to achieve extremely fast geometric pattern matching in two and three dimensional transformation space

We present a GPU-based approach to geometric pattern matching. We reduce this problem to finding the depth (maximally covered point) of an arrangement of polytopes in transformation space and describe hardware assisted (GPU) algorithms which exploit the available set of graphics operations to perform a fast rasterized depth computation. We give two alternatives, one is for translation + scale and the other is for rigid transformations, both have 3-parameters transformation space. We give extensive experimental results showing the running time of our method and its dependence on various parameters.