3-D object positioning from monocular image brightnesses
説明
This paper proposes an estimation scheme for three-dimensional (3-D) object positioning from monocular image .brightnesses without point correspondences while previous 3-D positioning methods need point correspondences. In the present scheme, at the first step, 3-D shapes of object are recovered from image brightnesses before and after rotation motion. At the second step, two vectors which are transformed according to rotation matrix under 3-D rotation, are constructed from central moments of the recovered 3-D shapes of object before and after motion, respectively. At the third step, rotation matrix is computed from these vectors by singular value decomposition. At the fourth step, from the rotation matrix the rotation axis and the rotation angle around the axis are estimated.