Research And Application Of Digital Image Inpainting In The Cracks Of Ancient Pictorial Carved Stone

The ancient pictorial carved stone has very precious historical and cultural values,and it is easily damaged which once the damage will not regenerate.With the application of computer image technology to the protection and inpainting of pictorial carved stone,the risk can be avoided,the interference of subjective factors in physical repair can be reduced,the process of stone protection and repairing can be promoted,and more convenient image of stone digital can be displayed.In this thesis,the stone image of Eastern King Father and Western Queen Mother?Guests Welcoming and Chest Playing are taken as examples.Due to the complexity in the cracks of pictorial carved stone,the result of classic inpainting algorithm is poor.Thus,the inpainting algorithm based on improved structure tensor and the Criminisi inpainting algorithm based on Markov random field matching criterion are researched.The result show that the inpainting accuracy of the stone image is improved.The main work is as follows:First,image inpainting algorithm based on improved structure tensor is researched.In the process of inpainting the damaged structure line in the stone image,for the unsmooth connecting of classical structure tensor with linear filtering,image inpainting algorithm based on improved structure tensor is researched.The nonlinear filtering constructs structure tensor instead of Gauss filter,and the improved nonlinear structure tensor is used to constrain the priority of inpainting algorithm.Finally,the structure information is diffused from the known region to the unknown region by calculating the direction and intensity of the edges.The experiment results show that the method can identify and inpaint the structure line.The inpainting result of the broken structure lines in pictorial carved stone image is nice in aspect of visual connectivity.Second,Criminisi inpainting algorithm based on markov random field matching criterion is proposed.The characteristics of the texture of the stone image is relatively smooth,uniform distribution,less energy.According to this,a new Criminisi digital inpainting algorithm based on Markov Random Field(MRF)matching criterion is proposed.And it is used to inpaint the broken texture in which image that the broken structure lines were already inpainted.The MRF becomes new matching criterion instead of Euclidean distance.Before searching the best matching patch,the image texture model based on MRF is built.Then the global energy of the image is calculated to estimate the value of inpainting patch.And the best matching patch is searched in order to achieve the global optimum.The experiment results show that the method achieved impressive inpainted results,the texture error matching rate is decreased,and the inpainting accuracy is improved.