Single-View Rolling-Shutter SfM

Sofía Errázuriz Muñoz; Kim Kiehn; Petr Hruby; Kathlén Kohn

cs.CVMar 12, 2026

Single-View Rolling-Shutter SfM

Authors: Sofía Errázuriz Muñoz, Kim Kiehn, Petr Hruby, Kathlén Kohn

Organizations: KTH Royal Institute of Technology, Sweden · Digital Futures, Sweden

Abstract

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.

Figures & tables

Panel of Figure 1 — Figure 1: Examples of RS images: left: iPhone 3GS sequence from [ 19 ] , middle : sequence 06 from [ 26 ] , right: synthetic image generated by [ 50 ] .

Panel of Figure 2 — Figure 2: Noiseless stability. Histogram of the velocity and line direction errors (for $\delta=0$ ) and the axis and norm errors (for $d=0$ ), calculated from $10^{5}$ noiseless samples.

Panel of Figure 3 — Figure 3: Noise test. Recall curve of the velocity and line direction errors (for $\delta=0$ ) and the axis and norm errors (for $d=0$ ), calculated from $10^{5}$ samples with noise $\sigma=1px$ , $||V||=0.2$ for $\delta=0$ , $||A||=0.2$ for $d=0$ .

Panel of Figure 4 — Figure 4: Real-world experiments. Recall curves of the velocity and line direction errors of the proposed solvers with $\delta=0$ on the dataset [ 26 ] ( left ), and of the axis and norm errors of the proposed solvers with $d=0$ on the dataset [ 19 ] .

Appendix figures & tables1 asset

Supplementary material from the paper’s appendix.

Appendix

Panel of Table 1 — Table 1: All minimal problems discovered in Section 4 . Each row contains all minimal problems in the given setting (except for $\delta=0,d=3$ ). The lines represent 3D lines, while the points are sampled on their image curves. The label identifies a problem as $dD\delta M(N)$ , where $D=\deg(C)$ , $M=\deg(A)$ , and $N=n_{1}^{e_{1}},\ldots,n_{k}^{e_{k}}$ means that $e_{i}$ lines have $n_{i}$ points sampled on their image curves. $P$ means that the lines are parallel and $PC$ means that they are parallel and coplanar. For the underlined labels, we evaluated minimal solvers in Section 6 .

Structure from Motion

Structure from Motion estimates camera poses and 3D scene geometry from overlapping images by matching visual features across views.

See all 46 papers

Rolling Shutter Relative Pose Estimation Made Practical
Jun 25, 2026Daniel BarathMulti-View Geometry Relative Pose Estimation
ETH Zürich, Computer Vision and Geometry Group · Google Zurich · HUN-REN SZTAKI, Budapest, Hungary
Rolling Shutter Camera Self-Calibration
Aug 2, 2026Yongcong Zhang, Navid Rabbani, Bangyan Liao +3Camera Calibration
DIA2M, DRCI, CHU Clermont-Ferrand, France · SURGAR, Surgical Augmented Reality, Clermont-Ferrand, France · School of Engineering, Westlake University, China +1
Estimating Velocity and Spin of Spherical Objects from Rolling-Shutter Image(s)
Jun 30, 2026Wenjie Xue, Jun Yang, Jingmin Wang +1
Epson Canada Ltd, Toronto, ON L3R 6G3, Canada · University of Toronto, Toronto, ON M5S 1A1, Canada

Jun 25, 2026J/K move · Enter open · S save

Rolling Shutter Relative Pose Estimation Made Practical

Daniel Barath

ETH Zürich, Computer Vision and Geometry Group · Google Zurich · HUN-REN SZTAKI, Budapest, Hungary

Multi-View Geometry Relative Pose Estimation Camera Pose Estimation Visual SLAM

Open paper AskPDF arXiv

Rolling shutter (RS) cameras equip virtually all consumer devices, yet RS-aware relative pose estimation has remained impractical: the state-of-the-art solver requires a minimum of 20 point correspondences, making RANSAC-based robust estimation prohibitively expensive due to the exponential dependence of the iteration count on the sample size. We make RS relative pose estimation practical by introducing affine correspondences (ACs) into the RS two-view geometry. We derive novel RS-corrected affine constraints that account for the coupling between point perturbations and the row-dependent essential matrix, providing two equations per correspondence beyond the standard epipolar constraint. Building on these constraints, we develop a linearized algebraic solver that estimates pose and RS motion from only 7 ACs. The solver exploits the physical smallness of RS parameters to linearize the constraints, eliminates the 12 RS unknowns via null-space projection, and solves the remaining degree-20 system via action matrices in 1.2,ms. On the TUM RS benchmark, our method achieves the best pose and RS parameter accuracy among all tested methods and, uniquely among RS solvers, provides accurate translational velocity estimates -- which are poorly conditioned from point correspondences alone due to a

\vec{v}

\vec{t}

coupling. On the global-shutter EuRoC MAV dataset, the solver achieves comparable accuracy to the standard 5-point algorithm, demonstrating that it generalizes well to the GS setting. Code is at https://github.com/danini/rolling_shutter_made_practical.

Abstract

Structure from Motion

Explore similar work

Rolling Shutter Relative Pose Estimation Made Practical

Rolling Shutter Camera Self-Calibration

Estimating Velocity and Spin of Spherical Objects from Rolling-Shutter Image(s)