C. Jonchery, G. Koepfler, and F. Dibos (France)
camera motion, parametric estimation, computer vision
In this paper, we propose a global approach for estimat ing the motion of a camera which films a static scene. The method is direct, fast and robust and deals with adjacent frames of a sequence. It is based on a quadratic approxi mation of the observed deformation between two adjacent images, in the case of a scene with constant depth in the camera coordinate system. This condition is very restric tive but we show that, provided translation and depth in verse variations are small enough, the error on the opti cal flow involved by the constant depths approximation is small. In this context, we propose a new model of cam era motion, that allows to separate the image deformation in a similarity and a “purely” projective application, due to change of optical axis direction. This model leads to a quadratic approximation of image deformation that we esti mate with an M-estimator; we can then immediatly deduce camera motion parameters.
Important Links:
Go Back
