Factorization methods for projective structure and motion

This paper describes a family of factorization-based techniques for the recovery of 3D scene structure and camera motion from multiple uncalibrated perspective images of 3D points and lines, up to an overall projective transformation. The methods can be viewed as generalizations of the Tomasi-Kanade algorithm from affine cameras to fully perspective ones, and from points to lines. They make no restrictive assumptions about scene or camera geometry, and unlike most existing reconstruction tech-niques they do not rely on ‘privileged ’ points or images. All of the available image data is used, and each feature in each image is treated uniformly. The key to projective factorization is the recovery of a consistent set of projective depths (pro-jective scale factors) for the image points. We show how this can be done using fundamental matrices and epipoles estimated from image measurements, and present a detailed study of the performance of the new reconstruction techniques as compared to several existing methods. We also describe an approximate structure/motion factorization technique that gives similar re-sults to Singular Value Decomposition based factorization, but runs much more quickly for large problems