Projective Reconstruction of Acquisition Geometry without Point Correspondences

Würfl T, Maass N, Dennerlein F, Maier A (2020)

Publication Language: English

Publication Type: Conference contribution, Conference Contribution

Publication year: 2020

Pages Range: 264 - 267

Conference Proceedings Title: Proceedings of the 6th International Conference on Image Formation in X-Ray Computed Tomography

Event location: Online meeting DE

Open Access Link: https://ct-meeting.org/?p=program

Abstract

Despite some fundamental differences in the image formation model 3D Reconstruction in the fields of Computed Tomography and Computer Vision share some similar concepts. In Computed Tomography, reconstruction refers to the problem of estimating 3D structure from 2D projection images assuming that accurate information about the camera geometry is available while in Computer Vision reconstruction typically refers to the task of simultaneously estimating the 3D structure and the camera motion solely from 2D projection images. In Computed Tomography accurate information about the camera geometry or motion is usually acquired by calibration and scanners are manufactured to support reproducible trajectories. Being able to reconstruct without requiring calibration and a reproducible trajectory would solve the problem of rigid motion estimation and enable new types of scanners. However, it is difficult to transfer Computer Vision techniques to Computed Tomography because point correspondences are hard to establish since all structures are superimposed over each other in projection images. Recently we introduced a new algorithm two estimate pairwise relative geometry based solely on the 2D projection images without point correspondences. In this work, we present the adaptation of a structure from motion algorithm to obtain a projective reconstruction of the whole scene from pairwise geometry estimates. In a simulation experiment, we show that we are able to estimate the camera geometry only based on projections.

Authors with CRIS profile

Tobias Würfl Lehrstuhl für Informatik 5 (Mustererkennung) Andreas Maier Lehrstuhl für Informatik 5 (Mustererkennung)

Involved external institutions

Siemens AG, Healthcare Sector DE Germany (DE)

How to cite

APA:

Würfl, T., Maass, N., Dennerlein, F., & Maier, A. (2020). Projective Reconstruction of Acquisition Geometry without Point Correspondences. In Proceedings of the 6th International Conference on Image Formation in X-Ray Computed Tomography (pp. 264 - 267). Online meeting, DE.

MLA:

Würfl, Tobias, et al. "Projective Reconstruction of Acquisition Geometry without Point Correspondences." Proceedings of the The 6th International Conference on Image Formation in X-Ray Computed Tomography, Online meeting 2020. 264 - 267.

BibTeX: Download