Mural Images Using Exemplar-based Inpainting Algorithm

Tang dynasty murals is the precious historical and cultural heritage for our country,butbecause of natural weathering and man-made destruction,lots of murals suffer fromdifferent degrees of damages.Image processing technology is applied to the muralprotection,it can imitate the mural restoration master manually repair process to getvirtual restoration of damaged murals information,this is a way to reduce the physicaldamage in the process of repair,and realize the mural can be non-contact restoration anddisplay.Digital image inpainting is one of the most important research content in imageprocessing area. Image inpainting is the process to reconstruct lost of deteriorated partsin images according to the mathematical model. With development of image processingand the machine vision, image inpainting is widely applied in the protection of culturalrelic and paintings, military remote sensing image processing, medical imageprocessing, image coding, video feature film effects production, image compression andso on.This paper overviews the status of research on Tang dynasty murals protection anddigital image inpainting techniques principles. An improved algorithm forexemplar-based image inpainting is proposed in this paper to solve the problems ofancient murals. The improved algorithm redefine the sampling way of sample piece,expands and gets new expansion of the sample searching space.The priority formula isimproved by containying the color information and effective area of sample,and thefilling strategy is improved according to the sample position from outside to inside,when a complete edge repair is completed, the new calculation.The experimental results confirmed that the algorithm improves the sample matching accuracy and reduces thecumulative error.On the basis of the algorithm, we build a smart and easy ancientmurals digital image inpainting system, it can identify annotation and automaticallyrepair a series of functions to the damage in the ancient murals,and support the ancientmurals digitized virtual exhibition.