Research On Digital Inpainting Models Of Collected Tomb Mural

Under the background of cultures digital subsistence,most of museums did high definition digitalization for their collections.It provided abundance of digital images that shows the information of culture relics hundreds years ago.This digital information is important for researching the history.It is a tendency to protect the relics using digital information recording before they fade away.Especially for the murals,if the painting on the surface is missing the mural will have nothing left.They are the most urgent collections need to be restored in the museum.However,there is always a long period before the murals can be restored,because the cumbersome process and understaffed of restorers.These murals will hind in the storeroom for a long time before the spectator can see them in the exhibition hall.The thesis research on the HD digital mural to simulate the restoration flow path and obtain the information left,the aim to build a digital inpainting system for museums to management.The Tomb mural digital inpainting is build based on Bayes inverse probability formula,it includes three parts as extract the prior knowledge,build data model and establish inpainting model.Considering the ill-posed inverse question,it should establish a non-reference objective evaluation abide by the Gestalt principle.In addition,the thesis designs TM-DIMP to adjust the digital inpainting system interactively using the experience of relic restoration experts.(1)Research on the prior knowledge.(1)Build the energy functional model constrained by edge to conduct the active contour snake model to contract subject and background of the murals.It can obtain the mural's main part layout information;(2)Choose the build data model by five color principle,and change the rule of differentiation from Euclidean distance to Bhattacharyya distance,and transform the K-means clustering analysis to Lab space to obtain the color layout information;(3)Design the Contour-let texture energy parameter to decompose the mural image into cartoon part and textural part with the frequency domain decomposition adaptively and obtain the structural and textural information.All these prior knowledge is important digital archives,and can guide the optimization of inpainting processing.(2)Research on the data model.Design multi-scale filtering on degraded tomb mural with the mathematical morphology by Top-hat and Below-hat,in order to highlight the image details of the damage areas.It applies the multi-scale structural elements to conduct the image detection for different disease types,thus obtaining the distribution diagram of three major diseases as crack,crevice and fracture,along with the damaged information of murals.All above statistics could serve as the mask in digital inpainting and guide the missing region convergence offering the data model for tomb mural inpainting.(3)Research on the inpainting model.(1)Build improved Curvature-Driven Diffusion model by changing the orthogonal spreading into cross diffusion,to adjust the diffusion process along with the direction of disease for inpainting in small-scale degrade areas;(2)Build improved exemplar-based model by changing the computing method of the priority value and adds the color item to reduce the weight of confidence item.And using the prior knowledge to setup the size of patches and the matching criterion to make better inpainting results in large-scale defect area;(3)Reduce the information redundancy with discrete sampling instead of traversing sampling,and enlarges the sampling database via multiple lens images,thus improving the inpainting efficiency and effects.(4)Stitched the detached blocks by Bezier curve to complete the structural of the losing part,and constrained exemplar inpainting between the new lines.Build the inpainting model to stitch for non-overlap areas of large tomb murals.(4)Establish the Tomb Mural Digital inpainting Management Platform(TM-DIMP)to join the online visitors,the remote experts,the inpainting technicians and museums managers.Abiding by the principle of Gestalt principle,this thesis designed a non-reference objective evaluation by means of block effect factors and fractal dimension to company with the subjective assessment,and implement the function of on-line preview,assignment collection,post feedback,interactive demonstration.This thesis proposed the Digital inpainting System for collected Tomb Mural.Using the HD digital mural data to extract the digital archives and rebuild the missing information in the murals.It can offer non-contact,undamaged,and repeatable restoration preview,to protect the mural from destroyed by blindness restoring,and shorten the period from storeroom to exhibition.It provides the basic theory research for development of mural digital information protection.