快速的自举鲁棒点匹配算法

鲁棒点匹配算法中的退火温度是按常数比率降低的,它导致运行时间长。文章基于对应关系矩阵的熵和退火温度是同时变化的现象,提出采用熵来建模退火温度,由此得到一种新的算法。尝试了两种实现方案,实验表明它们都比原算法高效,其中一种实现方案的鲁棒性和配准精度都非常接近原算法。

A Fast Bootstrap Robust Point Matching Algorithm

The annealing temperature in the robust point matching （RPM） method 1 ] is decreased by a fixed rate, which leads to long running time. Based on the observation that the entropy of correspondence matrix and annealing temperature change in accordance with each other, we propose in this paper using the entropy to model annealing temperature, which leads to a new point matching method. We tried two variants of this method. Experimental results showed that both of them are much more efficient than RPM, while one of them also well preserves the robustness and accuracy of RPM.