Search papers, labs, and topics across Lattice.
This paper characterizes the single-view geometry of rolling-shutter (RS) cameras for both points and lines, addressing a gap in RS structure-from-motion (SfM). By exploiting this geometry, the authors derive minimal reconstruction problems to determine which motion and scene parameters can be recovered from a single RS image. Proof-of-concept solvers are presented to evaluate the feasibility and limitations of the approach.
Unlock structure-from-motion from a single rolling-shutter image by exploiting its unique geometric properties.
Rolling-shutter (RS) cameras are ubiquitous, but RS SfM (structure-from-motion) has not been fully solved yet. This work suggests an approach to remedy this: We characterize RS single-view geometry of observed world points or lines. Exploiting this geometry, we describe which motion and scene parameters can be recovered from a single RS image and systematically derive minimal reconstruction problems. We evaluate several representative cases with proof-of-concept solvers, highlighting both feasibility and practical limitations.
