Lagrangian multipliers and split Bregman methods for minimization problems constrained on Sn-1

摘要: The numerical methods of total variation (TV) model for image denoising, especially Rudin-Osher-Fatemi (ROF) model, is widely studied in the literature. However, S^n^-^1 constrained counterpart less addressed. classical gradient descent method problem limited two aspects: one small time step size to ensure stability; other that data must be projected onto during evolution since unit norm constraint poorly satisfied. In order avoid these drawbacks, this paper, we propose alternative based on Lagrangian multipliers and split Bregman methods. Both algorithms are efficient easy implement. A number experiments demonstrate proposed quite effective denoising S^1 or S^2, including general direction diffusion chromaticity denoising.