Image Texture Smoothing Based On Local Mean Filter And Variational Regular Optimization Method

The key of the image texture smoothing technique is to remove the image texture and retain the significant edge structure of the image itself,which is widely used in the fields of image segmentation,edge detection,object recognition and image enhancement.Tradi-tional structure preserving filters and gradient regularization constraints are very effective in removing image noise,but they do not have the ability to distinguish image structures and textures.In recent years,the relative total variation technique,which is first explicitly given the measure function to distinguish the image structure and texture,has been further improved by many scholars at home and abroad because of its excellent image structure and texture recognition ability.It combines with the traditional structure preserving filter and the variational regular term constraint method and applied to the image texture smoothing technique.In this paper,the image texture smoothing technique is studied.The advantages and dis-advantages of the current texture smoothing method are analyzed and compared.A method based on rolling guidance propagation filtering and a method based on the minimization of₀norm based on window power difference are proposed to achieve better texture smoothing and structure retention.The main work of this article includes:Firstly,the current mainstream image texture smoothing methods are classified into two categories:local mean filtering and variational canonical optimization.The theoretical thinking and merits and demerits of some representative algorithms in each class are given.Secondly,this paper analyzes the shortcomings of some current local mean filtering methods in image texture smoothing,and proposes a method based on rolling guidance prop-agation filtering.In this method,we first use block translation and improved relative total variation technology to construct a new texture removal structure image.Then,the image is used as a guiding graph to conduct the rolling guided propagation filter,which realizes the structural restoration of images.The experimental results show that the proposed method can retain the significant edge structure of the image better than the traditional local mean filtering,while removing the image texture effectively.Finally,based on the advantages and disadvantages of the gradient₀norm minimiza-tion method and the relative total variation structure and texture separation method,a method based on the minimization of the₀norm of the window power difference is proposed,and the specific solution process is given.It first improves the original relative total variation measure,which satisfies the requirement of₀norm sparsity.Then the objective function is established,the₀norm of this function is minimized to get the optimal solution,and the ideal output image of the structure is obtained.The experimental results show that the proposed method can better maintain and sharpen the significant edge structure of the image,while effectively removing the texture,compared with the two methods mentioned above.