SVD in Fundamental matrix calculation
In Hartley's Geometry of Multiple View the follow method of Fundamental
Matrix calculation is proposed on p 281-282:Obtain correspondences,
normalize them, construct system of equation with dimension 9XN where N -
number of correspondences. The next step is to solve the system by SVD and
adjust the solution by constraint enforcement. Why the system should be
solved by SVD? Is there another way? I tried to solve the system with the
direct linear transform and got insufficient result. Whereas result
obtained by SVD is really good. (I checked it by epipolar constraint)
No comments:
Post a Comment