Scene planes and homographies Homographies given the plane and vice versa Proof of result 12.1 Example 12.2 A calibrated stereo rig A calibrated stereo rig 2 A calibrated stereo rig 3 The homography induced by a plane Fig.12.1 Fig 12.1 Legend Homographies compatible with epipolar Geometry Two sets of 4 arbitrary points from 2 images Epipolar geometry define conditions on homographies Counting degrees of freedom Compatibility constraints Fig.12.2 a e’ = H e Compatibility constraints 2 Fig. 12.2 b HT le’ = le Compatibility constraints 3 Fig. 12.2 c le' F x x, ( H x) Fig 12.2 Compatibility constraints Result 12.3 Homographies are compatible with fundamental matrix Corollary 12.4 Result 12.5 13.6 Plane induced homographies given F and image correspondences: (a) 3 points, (b) a line and a point 12.2.1 Three points Three points The first (explicit) method is preferred Degenerate geometry for an implicit computation of the homography Fig. 12.3 Fig. 12.3 Legend Determining the points Xi is not necessary in first method All that is important Result 12.6 Proof Proof 2 Consistency conditions Consistency conditions 2 Algorithms 12.1 The optimal estimate of homography induced by a plane defined by 3 points 12.2.2 A point and line A one parameter family of homographies Fig 12.4 (a), (b) Fig 12.4 Legend Result 12.7 Proof of result 12.7 Proof of result 12.7 (2) Proof of result 12.7 (3) Result 12.8 Result 12.8 2 Geometric interpretation of the point map H(m) Explore the A homography between corresponding line images Fig. 12.5 Fig. 12.5 Legend Degenerate homographies Degenerate homographies 2 A degenerate homography Fig. 12.6 Fig. 12.6 Legend 12.3 Computing F given the homography induced by a plane Plane induced parallax Plane induced parallax Fig. 12.7 Fig. 12.7 Legend Plane induced parallax 2 Fig. 12.8 Fig. 12.8 Legend Plane induced parallax 2 Algorithm 12.2 Computing F given the correspondence of 6 points, 4 of which are coplanar Fundamental matrix from 6 points of which 4 are coplanar Fig. 12.9 Fig. 12.9 Legend Projective Depth Example 12.9 Binary space partition: left and right images Fig. 12.10 a,b (c ) Points with known correspondence (d) A triplet of points selected from ( c ) and this triplet defines a plane Fig. 12.10 c,d (e) Points on one side of the plane (f) Points on the other side Fig. 12.10 e, f Fig 12.10 Legend Two planes Two planes 2 The action of the map H = H2-1 H1 on x Fig. 12.11 Fig. 12.11 Legend Two planes 3 Up to this points, the results of this chapter have been entirely projective 12.4 The infinite homography Hinf The infinite homography Hinf 2 The infinite homography Hinf 3 Vanishing points and lines The infinite homography Hinf maps vanishing points between images Fig. 12.12 Affine and metric reconstruction Affine and metric reconstruction 2 Affine and metric reconstruction 3 Stereo Correspondence Reducing the search region using Hinf Fig 12.13 Fig. 12.13 Legend