N.: Image uncertainty and pose estimation in 3d euclidian space

We describe a problem of a sucessful 3D{2D pose estimation algorithm when it is applied in scenarios with large depth variation. In this case image uncertainty is inhomomogenuously re ected in the Euclidian space where the constraint equations are formulated. We introduce a scaling of the constraint equations that equalizes this inhomogenity. We can show that we can reduce the error signicantly in outdoor scenarios with large depth discontinuities.