2008. 10. 26. 23:13
@GSMC/이상욱: Principles of Human and Artificial Vision
plane projective transformation
(Z_w = 0 -> 3-by-3 matrix representing a general plane to plane projective transformation)
http://en.wikipedia.org/wiki/Projective_plane
rotation about the optical center
: depth-independent transformation
(H does not depend on 3D structure.)
synthetic rotations
projective warping
http://en.wikipedia.org/wiki/Image_warping
null space
http://en.wikipedia.org/wiki/Kernel_(matrix)
The null space of a matrix with n columns is a linear subspace of n-dimensional Euclidean space.
The null space of A is the same as the solution set to the homogeneous system.
(cf. http://en.wikipedia.org/wiki/Kernel_(mathematics)
http://en.wikipedia.org/wiki/Kernel_(linear_operator) )
> 2d transformation
- Euclidian transformation
- affine transformation
http://en.wikipedia.org/wiki/Affine_transformation
- similarity transformation
http://en.wikipedia.org/wiki/Similarity_transformation
- projective transformation
http://en.wikipedia.org/wiki/Projective_transformation
> projective plane
homography
http://en.wikipedia.org/wiki/Homography
homogeneous coordinates
http://en.wikipedia.org/wiki/Homogeneous_coordinate
perspective projection
http://en.wikipedia.org/wiki/Perspective_projection#Perspective_projection
projective space
http://en.wikipedia.org/wiki/Projective_space
A point in the image is a ray in projective space.
=> All points on the ray are equivalent.
3D Projective Geometry
These concepts generalize naturally to 3D
- Homogeneous coordinates
Projective 3D points have four coords: X = (X,Y,Z,W)
- Duality
A plane L is also represented by a 4-vector
Points and planes are dual in 3D: LTP = 0
- Projective transformations
Represented by 4x4 matrices T: X’ = TX, X’ = L X -1
However
- Can’t use cross-products in 4D. We need new tools
Grassman-Cayley Algebra
: generalization of cross product, allows interactions between points, lines, and planes via “meet” and “join” operators
- Or just use inhomogeneous representation in 3D.
These concepts generalize naturally to 3D
- Homogeneous coordinates
Projective 3D points have four coords: X = (X,Y,Z,W)
- Duality
A plane L is also represented by a 4-vector
Points and planes are dual in 3D: LTP = 0
- Projective transformations
Represented by 4x4 matrices T: X’ = TX, X’ = L X -1
However
- Can’t use cross-products in 4D. We need new tools
Grassman-Cayley Algebra
: generalization of cross product, allows interactions between points, lines, and planes via “meet” and “join” operators
- Or just use inhomogeneous representation in 3D.
http://en.wikipedia.org/wiki/Grassman_algebra
vanishing point
http://mathdl.maa.org/mathDL/1//?pa=content&sa=viewDocument&nodeId=477&bodyId=598
vanishing line
'@GSMC > 이상욱: Principles of Human and Artificial Vision' 카테고리의 다른 글
Lecture 7: Image Filtering & Edge Detection (0) | 2008.10.28 |
---|---|
Lecture 6: Multiple-View Geometry: Stereo, Motion, Features (0) | 2008.10.27 |
Lecture 5: Single Camera View Geometry (0) | 2008.10.27 |
Lecture 3: Linear Algebra for Computer Vision (0) | 2008.10.26 |
Lecture 2: Image Formation & Camera Calibration (0) | 2008.10.24 |