Robust Rotation Synchronization via Low-rank and Sparse Matrix Decomposition

TL;DR

This paper introduces R-GoDec, a novel low-rank and sparse matrix decomposition method for robust rotation synchronization, effectively handling outliers and missing data in 3D registration tasks.

Contribution

It formulates rotation synchronization as a low-rank and sparse matrix decomposition problem and proposes the R-GoDec algorithm, which is faster and more robust than existing methods.

Findings

R-GoDec outperforms state-of-the-art algorithms in speed.

The method effectively handles outliers and missing data.

Experimental results validate its robustness and efficiency.

Abstract

This paper deals with the rotation synchronization problem, which arises in global registration of 3D point-sets and in structure from motion. The problem is formulated in an unprecedented way as a "low-rank and sparse" matrix decomposition that handles both outliers and missing data. A minimization strategy, dubbed R-GoDec, is also proposed and evaluated experimentally against state-of-the-art algorithms on simulated and real data. The results show that R-GoDec is the fastest among the robust algorithms.