Rotation-Only Bundle Adjustment
This work addresses the problem of improving rotation estimation accuracy for computer vision applications, particularly in multi-view settings.
This paper introduces a new method to estimate global camera rotations independently of camera positions and scene structure. By decoupling rotation estimation, the approach achieves consistent and significant accuracy gains when combined with state-of-the-art rotation averaging methods.
We propose a novel method for estimating the global rotations of the cameras independently of their positions and the scene structure. When two calibrated cameras observe five or more of the same points, their relative rotation can be recovered independently of the translation. We extend this idea to multiple views, thereby decoupling the rotation estimation from the translation and structure estimation. Our approach provides several benefits such as complete immunity to inaccurate translations and structure, and the accuracy improvement when used with rotation averaging. We perform extensive evaluations on both synthetic and real datasets, demonstrating consistent and significant gains in accuracy when used with the state-of-the-art rotation averaging method.