Wavelet-Based Multi-Scale Variation Image Noise Removal Model and the Image Geometry-Adapted Method for Multi-Scale Parameters Determining

In this paper, combining the two existing wavelet-based multi-scale variation (WBMV) models, wavelet-based multi-scale fourth-order partial differential equation (WBMF) and wavelet-based multi-scale total variation (WBMTV), we propose the wavelet-based multi-scale combination (WBMC) model, which overcome the defects of them. Furthermore, an image geometry-adapted (IGA) method for multi-scale parameters determining is proposed for WBMV model. The multi-scale parameters determined by IGA method (named IGA parameters) can radically promote the qualities of the restored image. In addition, we propose a multi-step time discretization scheme for WBMV model, which not only can accelerate the speed of convergence, but also can improve the qualities of the restored image. Numerical experiments show better performances of the proposed model, IGA method and multi-step time discretization scheme in preserving the shape of edges and important details, and removing the noise in image in comparison with the existing models and methods.