Robustness in motion averaging

摘要: The averaging of multiple pairwise relative motions in a sequence provides fast and accurate method camera motion estimation with wide range applications, including view registration, robotic path estimation, super-resolution. Since this approach involves the Lie-algebra underlying representation, it is non-robust susceptible to contamination due outliers individual motions. In paper, we introduce graph-based sampling scheme that efficiently remove such outliers. resulting global solution robust also an empirical estimate inherent statistical uncertainty. Example results are provided demonstrate efficacy our incorporating robustness averaging.