Reputation: 309
minimal set of correspondences for Essential matrix given intrinsics?
Fundamental matrix requires 8 correspondences. How many image-image correspondences are required for estimating essential matrix, given intrinsics and why? Since essential matrix is made out of Rotation and Translation, 3+2(translation is only upto a scale) there are 5 parameters that needs to be estimated, each correspondence gives one equation. Does it mean 5 correspondences are enough to estimate a essential matrix?
Upvotes: 0
Views: 578
Answers (1)
Reputation: 11785
Five point correspondences are enough to estimate the Essential Matrix. D. Nister provided an efficient solver back in 2004, immediately followed by a somewhat more elegant one by E. Stewenius. The Nister solver is the one used by the OpenCv "findEssentialMat" routine.
However, you have your math a bit wrong: 5 point correspondences in general position provide 10 linear constraints. If there were no ambiguities in the projection, a minimal solver would need 3 point correspondences.
Upvotes: 1
Related Questions
- Essential Matrix: Pure Rotation
- Is it possible to compute intrinsic and extrinsic camera parameters from a given camera projection matrix?
- Fundamental matrix correspondence equation results meaning
- Does multi plane calibration calculate only Intrinsic matrix?
- Calculating the essential matrix from two sets of corresponding points
- Camera pose estimation from essential matrix
- Find real world coordinate of a point from calibrated stereo cameras with fundamental and essential matrix
- Calculating essential matrix using rotation, translation and camera parameters
- determine camera rotation and translation matrix from essential matrix
- Essential Matrix from Fundamental Matrix in OpenCV