An efficient and locality-oriented Hausdorff distance algorithm: Proposal and analysis of paradigms and implementations
作者：
Highlights：
• I propose the first morphological implementation of the Hausdorff Distance (HD). Although Serra has proved that it can be computed using mathematics, he did not provide nor propose any tangible implementation. I am the first to propose a morphological implementation (there could be many) and I compare it using different paradigms (CPU and GPU) to the state-of-the-art.
• The proposed algorithm is much faster than the state-of-the-art (work of Taha and Hanbury published in TPAMI) when we consider both the CPU and the GPU for the worst case and for tasks such as image registration.
• I provide the most extensive evaluation of performance of the Hausdorff distance in the literature (considering different paradigms and programming languages).
• When we consider the CPU, my proposal is up to 8 times faster than the Taha-Hanbury implementation for image registration and 22337 times faster for the worst case. In the case of the GPU, it is even faster than that.
摘要
•I propose the first morphological implementation of the Hausdorff Distance (HD). Although Serra has proved that it can be computed using mathematics, he did not provide nor propose any tangible implementation. I am the first to propose a morphological implementation (there could be many) and I compare it using different paradigms (CPU and GPU) to the state-of-the-art.•The proposed algorithm is much faster than the state-of-the-art (work of Taha and Hanbury published in TPAMI) when we consider both the CPU and the GPU for the worst case and for tasks such as image registration.•I provide the most extensive evaluation of performance of the Hausdorff distance in the literature (considering different paradigms and programming languages).•When we consider the CPU, my proposal is up to 8 times faster than the Taha-Hanbury implementation for image registration and 22337 times faster for the worst case. In the case of the GPU, it is even faster than that.
论文关键词：Hausdorff distance,Mathematical morphology,Similarity,Registration,Parallelism
论文评审过程：Received 6 April 2020, Revised 24 January 2021, Accepted 5 April 2021, Available online 20 April 2021, Version of Record 15 May 2021.
论文官网地址：https://doi.org/10.1016/j.patcog.2021.107989
