Fast Gradient Projection Algorithm for Constrained Image Segmentation Model
Jiang Tao Cheng · Journal of Information Engineering University · 2011
This paper proposes a fast numerical algorithm for constrained convex image segmentation model.The algorithm combines the accelerating convergence strategy of fast iterative shrinkage/thresholding algorithm with gradient projection algorithm,and uses orthogonal projection to process the constraint conditions to avoid the irregularity of penalty term when penalty function method is employed.The algorithm also adopts dual method to avoid the nonsmooth and highly nonlinearity of total variation term.Numerical examples illustrate its validity and advantage of convergence speed.