Research On Murals Inpainting Method Based On CDD And Criminisi Models

Ancient murals,as a valuable historical relic,have always been regarded as a believable basis for studying the living customs of ancient people.However,there are usually cracks,shedding and mildew when unearthed because of the long term influence of the natural environment and man-made factors.These mural images can be restored by digital image processing technology.It can not only save costs,but also avoid the mistakes that may occur when manually repairing murals.Therefore,this paper uses the image processing technology to locate and restore the two common problems of cracks and shedding in murals.The main works are as follows:(1)Different location methods are applied for the two diseases of cracks and shedding in mural images.For the crack areas,the maximum entropy threshold segmentation and connected domain labeling algorithm are used to obtain a crack extraction map with the help of morphological processing to remove useless information.For the shedding areas,the preprocessing algorithm is determined through the study of several commonly used image graying and image enhancement methods,and the threshold segmentation and area growth algorithm are used in conjunction to complete the location of the shedding areas.(2)Through the analysis of the CDD(Curvature Driven Diffusions)algorithm model,it is found that the inpainting quality and efficiency are impacted when the curvature function and gradient value in the diffusion function are unreasonable in the inpainting process.The original algorithm is improved by setting the curvature adaptive function and adjustment factor.The improved algorithm can use different models automatically to implement restoration according to different curvature values.In addition,in order to improve the stability and rationality of the model,adjustment factor is introduced and the calculation method of gradient modulus is redefined.The improved CDD model is used to repair the cracked areas in the murals.(3)The shedding areas of murals are repaired by the improved Criminisi algorithm.Aiming at the problems of zero priority and single matching criterion in Criminisi algorithm,an improved scheme is proposed.According to the relationship between the feature value of the image structure tensor and the local pixel transformation,the local feature function of the image is constructed,which is introduced into the priority calculation formula of the original algorithm.In addition,the formula is changed to the form of segmented calculation.For the mismatching that may occur during the search,the Euclidean distance of the image is used as an additional condition during the algorithm search.As for the error accumulation problem during the repair,the update method of the confidence term is revised.(4)A digital restoration system for mural problems is designed and achieved.This system is developed using the GUI in Matlab and includes four modules:crack location and repair,shedding location and repair,repair effect improvement and storage.