Abstract：Image noise and outlier always exist in the estimation of fundamental matrix.As a result,a new method is proposed based on projective space.First,a special fundamental matrix is defined including only 5 parameters according to the projective standard basis of three-view geometry.The method can cope with mismatch and search for the best 7 correspondence points to estimate epipolar geometry based on RANSAC(random sampling consensus)technique according to the principle of the minimal re-projection error.Many experiments with both synthetic and real images show that our method can obtain fundamental matrix with high accuracy.
 Hartley R. In defense of the eight-point algorithm [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1997, 19(6):580-593.  Zhang Z Y. Determining the epipolar geometry and its uncertainty:a review[J]. International Journal of Computer Vision, 1998, 27(2):161-198.  Torr P H S, Murray D W. The development and comparison of robust methods for estimating the fundamental matrix[J]. International Journal of Computer Vision, 1997, 24(3):271-300.  Zhang Z Y, Deriche R, Faugeras L, et al. A robust technique for matching two uncalibrated images through the recovery of the unknown epipolar geometry[J]. Artificial Intelligence, 1995, 78(1-2):87-119.  Quan L. Invariants of 6 points from 3 uncalibrated images[A]. The Third European Conferce on Computer Vision [C]. Germany:Springer-Verlag, 1994. 459-469.  Xiong Y, Peng J X. The unique solution of projective invariants of six points from four uncalibrated images[J]. Pattsrn Recognition, 1997,30(3):513-517.  Hartley R I, Dano N Y. Reconstruction from six-point sequences[A]. IEEE Computer Society Conference on Computer Vision and Pattern Recognition[C]. USA:IEEE Computer Society, 2000. 480-486.  Lei C, Tsui H T, Hu Z Y. On the automatic estimation of fundamental matrix[A]. The 5th Asian Conference on Computer Vision[C].Melbourne, Australia:Springer, 2002. 143-150.  Schaffalitzky F, Zisserman A, Hartley R I, et al. A six point solution for structure and motion[A]. The 6th European Conference on Computer Vision[C]. Dublin, Ireland:Springer, 2000. 632-648.  Harris C, Stephens M. A combined corner and edge detector[A].Proceedings of the Fourth Alvey Vision Conference[C]. Manchester:1988. 189-192.