| Image denoising is one of the most important tasks, which can be implemented efficiently using variational models for image diffusion. The classic TV(Total Variation) model can be used to diffuse images with edge preserving, but the computation efficiency is low which results in considerable attention in image processing area. Some improvements and applications are presented in this paper, which include: Implementations of TV-L~2,TV-L~1 models are given using 4 points neighborhood and 8 points neighborhood. A new median formula is proposed for higher order image diffusion based on 4 points neighborhood and TV-L~2,TV-L~1 models to reduce staircasing effects. The median formula is used to complex computation of multiscale denoising model to improve computing efficiency. Numerical examples are given to validate the formulas suggested in this paper. |
