Motion estimation between two views especially from a minimal set of features is a central problem in computer vision. Although algorithms for computing the motion between two calibrated cameras using point features exist, there is up to now no solution for a scenario employing just circles in space. For this task a minimal and linear solver is presented using only two arbitrary circles. It is shown that the derived algorithm is not hampered by critical configurations. Here, the problem of relative pose estimation is cast into a combined reconstruction plus absolute orientation algorithm. Thus, even for small or vanishing translations the relative pose between cameras can be computed in a stable manner within a single algorithm.
展开▼
