Research On Image Inpainting Based On C-H Equation And Texture Synthesis

With the development of information technology in today's society,image inpainting as a hot research direction in digital image processing technology,has been widely applied,involving medicine,cultural relic protection,film and television industry and so on.In the process of image collection and storage,some information may be contaminated.Image repair is to fill the damaged area according to the known part of the image,restore the intact image,and make the image conform to the visual rationality of human eyes.Image inpainting algorithms can be divided into three categories: image inpainting algorithms based on partial differential equation,image inpainting algorithms based on texture synthesis and image inpainting algorithms based on image decomposition.This paper studies these three kinds of image inpainting algorithms and proposes improvement methods respectively.The main research work of this paper is as follows:In terms of image inpainting based on partial differential equation,the main research Cahn Hilliard equation model(C-H),first put forward using fourth-order nonlinear diffusion term instead of biharmonic item,then using histogram to construct many potential well function,double potential well replace function,income model will be used to repair the gray image,the numerical calculation of the model based on Fourier and convex splitting method.Experiments show that the model is effective in repairing large area scratch occlusion in binary image and gray image.In the aspect of image inpainting based on texture synthesis,the Criminisi algorithm is mainly studied.For the priority calculation,the correlation between the sample block to be repaired and the neighborhood block is added to make the priority calculation more reasonable.For the selection of sample block,the adaptive selection of sample block size is used to replace the original fixed size to improve the repair efficiency.For the matching criterion of the sample block,relevant parameters in the HSV color space are added to calculate the square sum of the difference of corresponding pixel parameters to judge the similarity,which can better repair the color image.In terms of image inpainting based on image decomposition,the image using the decomposition model is decomposed into structure and texture map,respectively proposedtwo kinds of improved algorithm is used to repair,the structural part using the modified C-H equation model to repair,and texture part adopts improved Criminisi algorithm,combining repair good structure parts and texture,the experimental results show that with the C-H equation and Criminisi algorithm based on improved repair effect is better.