In this paper, a new approach is proposed to solve the problem of 3D reconstruction of scenes from a single image taken by a light-field camera equipped with a rolling shutter sensor. The method utilizes 3D information cues in the light-field and motion information provided by the rolling shutter effect. A two-stage algorithm that minimizes the re-projection error while considering the position and motion of the camera is also introduced. The researchers also present a new benchmark dataset of different light-fields showing rolling shutter effects. The effectiveness and advantages of the approach are demonstrated through several experiments conducted for different scenes and types of motions.
Publication date: 3 Nov 2023
Project Page: https://github.com/ICB-Vision-AI/RSLF
Paper: https://arxiv.org/pdf/2311.01292
