Structure from Motion for Panorama-Style Videos (1906.03539v1)

Abstract: We present a novel Structure from Motion pipeline that is capable of reconstructing accurate camera poses for panorama-style video capture without prior camera intrinsic calibration. While panorama-style capture is common and convenient, previous reconstruction methods fail to obtain accurate reconstructions due to the rotation-dominant motion and small baseline between views. Our method is built on the assumption that the camera motion approximately corresponds to motion on a sphere, and we introduce three novel relative pose methods to estimate the fundamental matrix and camera distortion for spherical motion. These solvers are efficient and robust, and provide an excellent initialization for bundle adjustment. A soft prior on the camera poses is used to discourage large deviations from the spherical motion assumption when performing bundle adjustment, which allows cameras to remain properly constrained for optimization in the absence of well-triangulated 3D points. To validate the effectiveness of the proposed method we evaluate our approach on both synthetic and real-world data, and demonstrate that camera poses are accurate enough for multiview stereo.

The paper entitled "Structure from Motion for Panorama-Style Videos" introduces a novel pipeline designed to accurately reconstruct camera poses for videos captured in a panorama style, even without prior camera intrinsic calibration. This method addresses the unique challenges posed by panorama-style video capture, characterized by rotation-dominant motion and small baseline between views, which traditionally result in inaccurate reconstructions.

Key contributions of the paper include:

1. Spherical Motion Assumption: The authors propose that the camera motion can be approximated as occurring on a sphere. By leveraging this assumption, they are able to introduce specialized relative pose estimation methods that are tailored to spherical motion, enhancing the overall accuracy of the reconstruction.
2. Novel Relative Pose Methods: They develop three novel solvers to estimate the fundamental matrix and account for camera distortion within the context of spherical motion. These solvers are designed to be both efficient and robust, providing a strong initialization for the subsequent bundle adjustment step.
3. Soft Prior for Bundle Adjustment: During the bundle adjustment process, they incorporate a soft prior that discourages significant deviations from the spherical motion model. This ensures that the camera poses remain properly constrained, which is crucial in scenarios where well-triangulated 3D points are sparse.
4. Validation and Effectiveness: The proposed method is thoroughly validated using both synthetic and real-world data. The results demonstrate that their approach yields highly accurate camera poses, suitable for applications such as multiview stereo.

By introducing an innovative Structure from Motion pipeline with a specialized approach for panorama-style videos, this paper makes a significant contribution to the field of computer vision. The ability to achieve accurate camera pose reconstruction without prior intrinsic calibration extends the usability of Structure from Motion techniques to a wider range of video capture scenarios.