In the general case, a trilinear relationship between three perspective views is shown to exist. The trilinearity result is shown to be of much practical use in visual recognition by alignment—yielding a direct reprojection method that cuts through the computations of camera transformation, scene structure and epipolar geometry. Moreover, the direct method is linear and sets a new lower theoretical bound on the minimal number of points that are required for a linear solution for the task of reprojection. The proof of the central result may be of further interest as it demonstrates certain regularities across homographies of the plane and introduces new view invariants. Experiments on simulated and real image data were conducted, including a comparative analysis with epipolar intersection and the linear combination methods, with results indicating a greater degree of robustness in practice and a higher level of performance in reprojection tasks.
|Original language||American English|
|Number of pages||11|
|Journal||IEEE Transactions on Pattern Analysis and Machine Intelligence|
|State||Published - Aug 1995|
Bibliographical noteFunding Information:
I acknowledge Office of Naval Research grants N00014-92-5-1879 and N00014-93-1-0385, National Science Foundation grant ASC-9217041, and ARPA grant NOOO14-91-J-4038 as sources of funding for the Artificial Intelligence Laboratory and for the Center for Biological Computational Learning. Also acknowledged is the McDonnell-Pew postdoctoral fellowship that has been my direct source of funding for the duration of this work. I thank T. Luong and L. Quan for providing their implementation for recovering fundamental matrices and epipoles. Thanks to N. Navab and A. Azarbayejani for assistance in capturing the image sequence (equipment courtesy of MIT Media Laboratory).
- Visual recognition
- algebraic and geometric invariants
- projective geometry