Recovering Multiple View Geometry from Mutual Projections of Multiple Cameras

摘要

In this paper, we analyze the computation of epipolar geometry in some special cases where multiple cameras are projected each other in their images. In such cases, epipoles can be obtained directly from images as the projection of cameras. As the result, the epipolar geometry can be computed from less image correspondences with higher stability. In this paper, we analyze the number of corresponding points required for computing bifocal, trifocal and quadrifocal tensors linearly in the case where cameras are projected mutually. We next show a practical linear method for computing multifocal tensors by using the mutual projection of cameras. The degenerate configurations of points and cameras is also studied, and it is shown that some degenerate configurations in general cases are no longer degenerate under the mutual projection of cameras.