Dual Computation of Projective Shape and Camera Positions from Multiple Images

摘要: Given multiple image data from a set of points in 3D, there are two fundamental questions that can be addressed: What is the structure 3D? What positions cameras relative to points? In this paper we show that, for projective views and with position defined projectively, these problems dual because they solved using constraint equations where space camera occur reciprocal way. More specifically, by canonical reference frames all images, imaging point sets captured relations involving three different kinds parameters only, coordinates of: (1) points, (2) (3) points. The duality implies problem computing fromp q same algorithm as directly reconstructing q+4 p-4 views. This unifies approaches reconstruction: methods based on external calibration direct exploiting constraints exist between shape invariants.