3D non-rigid motion estimation using the improved simulated annealing algorithm

An approach to 3D non-rigid motion estimation using improved simulated annealing algorithm (SA) from monocular image sequence is proposed. First, the affine motion model of non-rigid object is presented. Then, the energy function, which can reflect the joint probability distribution of motion parameters and the constraints relation of local 3D motion parameters, is established in Markov random field (MRF). After the limitations of classical SA and the features of motion-parameters estimation are analyzed, an improved SA is used to minimize the energy function. Experiment results from two synthetic image sequences demonstrate the robustness of solutions and the feasibility of our algorithm.