Research On Intelligent Restoration And Optimization Methods Of Ancient Mural Images Based On Generative Adversarial Network

Ancient Chinese murals have a long history and a large number of types.They are a testimony to the development of ancient Chinese civilization.They still have important historical,artistic,scientific,cultural relics,economics,missions and other values.In recent years,the application of scientific and technological means has played an important role in the protection of ancient murals,making the working methods of cultural relic protection more scientific and diversified.In the context of increasingly rich digital protection methods for cultural relics,the protection of murals requires more innovative work.This article takes ancient Chinese murals as the research object,and conducts research on the phenomenon of mural damage and color blurring.First,use the method of enhancing the overall details of the image to repair the damaged ancient mural image,and then use the super-resolution reconstruction method to optimize the ancient mural image,making the mural image more beautiful and artistic,and providing protection for the mural new ideas.First of all,in view of the difficulty of collecting the image data of ancient murals,the original mural images will be obtained by crawling technology and on-site shooting methods,and the mural images with severe damage,low resolution,and more interference information will be filtered out,and the mural images that meet the experimental conditions will be selected.As the initial data,the data set is constructed by the data augmentation algorithm and used as the research basis.Then,with the generative adversarial network as the framework,the masked mural image is completed in the generation network,and the discrimination network judges the output image of the generation network and the real mural image,and gradually completes the generation network during the confrontation.Modeling,until the model achieves the expected repair effect,complete the initial repair of the mural image.Then,the convolutional neural network is used to extract the feature information of the mural image,and then the complex feature information is mapped to a specific size of the image feature space to complete the super-resolution reconstruction of the mural image.Finally,build a mural image management prototype system to complete the collection of mural images more scientifically,and at the same time use the mural image superresolution reconstruction algorithm as the core to visually complete the mural image super-resolution reconstruction work.This thesis has completed the construction of ancient mural image data set,the research of mural image restoration method,the research of mural image super-resolution reconstruction method,and the development of the mural image management prototype system.This article not only repairs the mural image,but also uses super-resolution reconstruction technology to optimize the mural image.In the construction of the data set,the use of web crawler technology,data enhancement and other methods has laid the foundation for the application of deep learning in the field of mural protection in the future.And in the network training process,other network structure optimization methods such as transfer learning ideas are used to optimize the mural image,which provides an important reference for the application of deep learning in the scene of small sample data sets.