Split The Two-phase Moving Object Segmentation Method

Optical computation is widely used in many fields, for example, computer vision, three-dimensional reconstruction and so on. There are many ways to calculate the optical flow field, the variational method is an effective method to calculate the dense optical flow field, but the calculation costs long time and the efficiency is low.Aiming at the models with robust data terms and smooth terms in TV(Total Variation) type. On one hand, the split method for high computation efficiency is designed, the original minimization problems are split to some sub-minimization problems through introducing auxiliary variables which can be implemented by soft thresholding formulas, shrinking operators and Gauss-Seidel iterations. The original variation method and the split method are compared for data terms with constant assumption of intensity, gradient, Hessian matrix, norm of gradient, Laplacian and determinant of Hessian matrix. On the other hand, the Split Bregman algorithm is designed through introducing auxiliary variables and iterative parameters which achieve the fast optical flow calculation, and the Split Bregman method is compared with the original variation method.In the processing of video, the motion segmentation plays an important meaning. Although the original variational method can be used, it is not immediate, this split method is applied to the two phase motion segmentation model and is compared with the original variational method, results prove that split method is effective to improve the speed of computation.The rapid split method is studied in aspects of optical computation and two phase motion segmentation, it gets some valuable results. The split method is still a potential area of research.