Multivariate Barzilai-Borwein Method and Its Application in Elastic Registration of Medical Image

Combined with the advantages of multivariate spectral gradient method and spectral gradient method,multivariate Barzilai-Borwein method is proposed for unconstrained optimization.Moreover,based on a nonmonotone line search,global convergence is established for multivariate Barzilai-Borwein algorithm,and the algorithm is applied the B-spline parametric models in medical image elastic registration.The parametric models are often transformed into unconstrained optimization problem on image of multi-parameter,but the traditional methods are time-consuming for the problem with a large number of parameters.Multivariate Barzilai-Borwein algorithm utilizes more gradient information and greatly enhances the speed and precision of optimization in image registration,at the same time muti-resolution B-splines and the nonmonotone line search contribute to avoid local optimal.