Information Restoration Of Sootiness Area In Temple Murals Based On Spectral Imaging

The changes of environment and human activities will lead to the serious degradation of murals.Sootiness is one of the most common diseases in ancient Chinese murals.Hyperspectral imaging technology can obtain hyperspectral data of murals in the range of visible light to near infrared light,which provides a scientific and non-destructive method for digital protection of murals.In view of the problem of information restoration in the sootiness area of temple murals,this paper makes full use of the advantages of abundant bands of hyperspectral images and stronger penetration ability of near infrared bands than visible light to realized the image enhancement and information extraction of the heavy sootiness mural images.The restoration of light sootiness mural image was realized by introducing the relevant image defogging method.The virtual restoration of the paint losses diseases was carried out by using the image inpainting algorithm,which further improved the visual effect of the mural.The main research results are as follows:(1)Proposing the image enhancement and information extraction method of heavy sootiness mural image.For the heavy sootiness mural hyperspectral image which is seriously polluted by sootiness and the pattern is difficult to identify.Firstly,the image noise is limited and the processing accuracy is improved by radiometric correction,target area clipping and data denoising preprocessing.Then,the minimum noise fraction rotation is used to reduce the sootiness and scratch characteristics in murals,and enhance the pattern information.Through analyzing the average spectral curves of the background and pattern of the mural,the feature band with the largest difference is selected for image subtraction to enhance the mural pattern.Finally,the enhanced image is obtained by reflectance transform,and the pattern information under sootiness coverage is extracted by density segmentation and morphological filtering,which increases the readability and artistic expression of murals.The results show that the proposed method can extract the image information of the heavy sootiness mural.By comparing with the common image enhancement method and blind separation technology,it is shown that the proposed method is good for improving the visual quality of the heavy sootiness mural image.(2)Proposing the image restoration method of light sootiness mural image.For the light sootiness mural hyperspectral image with light sootiness degree and dark picture as a whole.Firstly,through radiometric correction,band clipping and minimum noise fraction rotation,the denoised reflection image is obtained.Secondly,according to the advantage that the near-infrared band can provide information under the surface material coverage to a certain extent,which is helpful to sootiness removal,the near-infrared band is selected from the reflection image,and the green and blue visible light bands are used to synthesize the pseudo color image,and the block histogram matching of the pseudo color image is performed based on the true color image to obtain the sootiness mural image,which can be used for subsequent sootiness removal restoration of murals.Thirdly,based on the atmospheric scattering model,the dark channel prior method is used to remove the sootiness in the mural image.Finally,according to Retinex method,the sootiness mural image is restored by using two bilateral filters.The results show that,compared with homomorphic filtering and Gaussian stretching,the proposed method has obvious advantages in variance,average gradient,information entropy and gray contrast,and has the highest score in the comprehensive evaluation of edge,tone and structure.Therefore,the proposed method can effectively reduce the influence of sootiness on mural patterns and improve the visual quality of murals.(3)Proposing the inpainting method for paint losses diseases.For the paint losses diseases of the light sootiness mural image background.Firstly,the Support Vector Machine(SVM)and morphological filtering is used to extract and expand the area of paint losses diseases in mural background to get the area to be inpaited.Then,the initial sootiness removal image is treated by Criminis algorithm.Finally,the restoration of sootiness murals is carried out by the method of Retinex by bilateral filter,and the images after restoration are mosaicked to one image,which can further improve the visual effect of mural restoration.The results show that the method can achieve the virtual restoration of paint losses diseases and sootiness mural restoration,and reproduce the original appearance and original style of murals.Therefore,this study realizes the information restoration of heavy and light sootiness area of temple murals,improves the connotation and artistic expression effect of ancient murals,increases its circulation and display ability,and provides scientific reference for its protection,restoration and research.At the same time,there are some contents that can be further studied or improved in this paper:(1)Because the spectral curves of different substances are different,only a single background pattern is enhanced and extracted in the process of image enhancement and information extraction of heavy sootiness mural.We should consider how to achieve visual enhancement and information extraction when there are many substances in the image or the spectral difference between the target and other substances is small.(2)Although the formation and manifestation of sootiness are similar to that of fog,there are still some differences.Therefore,it is necessary to further study the radiation transmission model which is more suitable for sootiness murals,so as to realize the restoration of sootiness murals more effectively.In addition,how to restore the rich color and complex texture patterns in murals robustly,so as to better restore the original appearance of murals,is also worthy of in-depth study.