Definition
SVD null-space estimation is the standard technique for finding a geometric entity that is defined only up to scale from a set of homogeneous linear constraints. Given a coefficient matrix assembled from measurements, it returns the unit vector that minimises subject to .
Let be a design matrix () built by stacking one linear constraint per measurement into the homogeneous system . The unique minimiser of subject to is
the last column of in the SVD , i.e. the right-singular vector corresponding to the smallest singular value .
The input is — any coefficient matrix whose rows encode linear constraints on an unknown vector meaningful only up to a global scale. The output is the unit vector that best satisfies those constraints in the least-squares sense. In computer vision is called the design matrix or DLT matrix; the pattern of stacking one constraint per correspondence and extracting the smallest right-singular vector appears identically in fundamental-matrix estimation, homography estimation, and camera calibration.
Mathematical Description
The homogeneous system and its least-squares solution
A geometric entity described by parameters but defined only up to scale has effective degrees of freedom. Each measurement contributes one linear equation; stacking constraints yields
which has no non-trivial solution unless is exactly rank-deficient. Under measurement noise has full column rank, and the exact equation is replaced by the constrained minimisation . Expanding via the SVD with ,
Subject to , this weighted sum of squared projections is minimised by placing all weight on the smallest term — — giving minimum value . This is the Rayleigh-quotient argument applied to , whose eigenvalues are and eigenvectors the columns of .
Rank deficiency and the residual interpretation
An exact solution to exists if and only if : the system is rank-deficient and lies exactly in the null space. When , no exact solution exists and is the minimum achievable residual . The ratio measures how clearly the solution direction is separated: a large ratio indicates a well-determined unique solution; a ratio near 1 indicates near-degeneracy.
For the eight-point fundamental-matrix estimator, is the design matrix from correspondences, each row ; the nine-vector reshapes to satisfying . The stacking follows from the bilinear epipolar constraint that Longuet-Higgins derived in 1981, where each correspondence supplies one equation in the nine entries of the essential matrix. For the homography DLT the same structure applies with a matrix; for Zhang's planar calibration the stacked system is a matrix whose smallest right-singular vector encodes the image of the absolute conic.
Rank enforcement by truncating the spectrum
The null-space step produces an unconstrained solution. Many geometric entities carry an additional rank constraint, enforced by a second SVD applied to the recovered matrix. The fundamental matrix must have rank 2, yet the linear solution generically has rank 3. The closest rank-2 matrix in Frobenius norm is obtained by computing and setting — zeroing the smallest singular value. This projection is optimal in Frobenius norm but not in geometric (Sampson) error; iterative refinement is needed for calibration-grade accuracy.
Numerical Concerns
Conditioning of the design matrix. The fundamental-matrix design matrix has entries ranging from to for 1000-pixel coordinates, making the condition number of of order – on un-normalised data. At such conditioning, 64-bit floating point retains only a few significant digits in the smallest singular vector. The fix — dlt-normalisation — translates each point set to zero centroid and scales isotropically to unit mean distance, dropping the condition number by roughly and bringing all design-matrix entries to .
The singular-value gap as a degeneracy indicator. A ratio near 1 means two or more directions of compete for the null space — the solution is numerically unreliable and geometrically degenerate. For the eight-point algorithm this corresponds to degenerate correspondence configurations such as collinear or coplanar points; for Zhang's system, parallel-plane views contribute linearly dependent rows.
Exact vs overdetermined systems. With exactly consistent constraints, and is the unique null vector. With constraints the system is overdetermined, , and is the least-squares null vector. The eight-point algorithm is most stable from roughly ten correspondences onward.
Sensitivity to outliers. The null-space step minimises a global least-squares objective over all rows of ; a single grossly mismatched correspondence corrupts the entire fit by inflating the residual in a direction unrelated to the true null vector. The method has no intrinsic outlier rejection — in practice RANSAC wraps it: sample a minimal subset, compute , count inliers, repeat, then re-estimate from the inlier set.
Precision. Applying SVD directly to is numerically preferable to forming and computing its eigendecomposition, since squaring the matrix squares the condition number. Even after normalisation, the smallest singular value and its right-singular vector should be computed in 64-bit double precision.
Where it appears
The SVD null-space step is the shared computational core of every homogeneous linear estimator in the atlas; dlt-normalisation describes the preconditioning that must precede it.
- longuet-higgins-eight-point — the 1981 origin: eight correspondences determine the nine entries of the essential matrix up to scale via SVD of the design matrix.
- fundamental-matrix-eight-point — Hartley's normalised extension; SVD of the matrix yields , followed by a second SVD enforcing rank 2.
- gao-dual-homography-stitching — fits one homography per correspondence cluster, each solved by SVD null-space estimation from its cluster's DLT design matrix.
- homography — the projective planar mapping; the DLT matrix assembles two rows per correspondence and the nine-vector reshapes to the homography.
References
- H. C. Longuet-Higgins. A computer algorithm for reconstructing a scene from two projections. Nature, 293–135, 1981.
- R. I. Hartley. In Defense of the Eight-Point Algorithm. IEEE Transactions on Pattern Analysis and Machine Intelligence, 19(6)–593, 1997.
- Z. Zhang. A Flexible New Technique for Camera Calibration. IEEE Transactions on Pattern Analysis and Machine Intelligence, 22(11)–1334, 2000.
- J. Gao, S. J. Kim, M. S. Brown. Constructing Image Panoramas Using Dual-Homography Warping. IEEE CVPR, 2011.
- R. Hartley, A. Zisserman. Multiple View Geometry in Computer Vision, 2nd ed. Cambridge University Press, 2004.