| Image processing is a science with rapid development, which has wideapplication in many fields. The progress of the computer science and the popularityof image display device become the main power of promoting image processing,which also provide a good external conditions.In the past ten years, image processing based on PDE has got rapiddevelopment, which has become one of the methods with more effect in the field ofimage processing. PDE has the perfect theoretical basis and wide range of numericalmethod. Previous put forward many models and algorithms using these favorabletool. The traditional image denoising method is based on the linear system, e.g.Winner Filter etc. These methods can only handle a simple question, but it’s notapplicable to large matrix condition number. Rudin, Osher and Fatemi present theROF model,which improve the condition number to some extent, along the regularfilter. However, the ROF model has the serious nonlinearity.2007, the splitBregman iteration method which is presented by Goldstein and Osher has beenproved to be the one of the most efficient way to deal with the aforementionedproblems. The basic idea of this model is construct the equality constraint, which isto replace the complex element (u)in L1item with an easy function. Using thesystemizing, put uinto the quadratic term of differentiable.In this paper, I mean to join two relaxation factors into the process of the twoiterations in the Split Bregman iterative algorithm respectively. So can we obtain thenew double relaxation split Bregman iteration algorithm (DRSB). This way is usedto process the anisotropic total-variation denoising of an image. Showing from theexperiment, compared with the original algorithm, through adjusting the tworelaxation factors, the improved algorithm has the faster convergence speed andbetter denoising effect. What’s more, in the process of selection and analysis of relaxation factor, we have the best selection of relaxation factor range. |
PDF Full Text Request