Research Of Color Restoration Technology For Fading Cultural Relic Model

The surface texture of the excavated cultural relics has been faded seriously by weathering oxidation, the color texture reconstruction is the important research topic of the digital protection of cultural relics and has been widely used in the fields of virtual tourism, cultural industry and digital archives of cultural relic. However, the color texture reconstruction technology for the fading entity has not been consummated. In this thesis, the key techniques of color texture mapping, color restoration and multi view texture mapping are studied, and the solution method of color texture reconstruction is established. In the method, to improve the diversity effect of virtual display of cultural relics, the surface texture image of color fading model are get from multi-view, and color restoration based on expert sample and residual color, then 2D-3D texture mapping is set up for the model of color restoration and reconstruction of faded cultural relics. The main research work and innovations are as follows:(1) A non-uniform and multi-view texture acquisition method based on the texture distribution of fading cultural relics is proposed. For the problem of quantitative calculation of non-uniform texture acquisition viewpoint, camera calibration is used to design acquisition environment, and the division of viewpoint is established for non-uniform rational allocation of collection viewpoint according to the model distribution of color number and geometric patterns,. Viewpoints in the texture dense region are more than the viewpoints in the texture sparse region. Experiments show that the accuracy of the measurement of quantitative non-uniform is improved.(2) A color transfer method based on independent component analysis (ICA) and mixed color distance is proposed. To solve the problem of error matching of the traditional color transfer -Welsh method caused by the high consistency requirement of color and brightness. According to the gray scale distribution of the faded image, using ICA method to extract independent components in order to reduce the consistency, improve the matching accuracy. The color distance similarity function is improved based on luminance distance and neighborhood standard deviation to measure the color distance between the color reference image and restored gray image. It can effectively improve the similarity degree of color matching.(3) A color restoration method based on combined SVM regression is proposed. To solve the problem of similarity comparison for color transfer in the gray scale distribution smoothly faded image, two kinds of regression of the channel 1 respectively to ? and ? in color space 1?? are utilized in the recolor problem in order to establish the corresponding SVM regression model, and the two models are combined for gray scale distribution smoothly image to estimate the color pattern. By analysing three errors index of SMAPE, RMSE and MAPE, combined SVM color restoration method is more accurate.(4) A color restoration image quality evaluation method based on human visual characteristics is proposed. Combining Peak Signal to Noise Ratio (PSNR) and human visual feature information entropy, a weighted quality evaluation function is established to compare the correlation degree of the restored color image and the reference color image, and objective judgment the quality of the color restoration method. The evaluation results were consistent with the subjective perception of human vision.(5) A 2D-3D texture mapping method based on multi-view salient feature points is proposed. To solve the problem of the automatic detection of feature points in color restoration of multi-view texture image and geometric model matching, the more feature points are utilized in non-uniform multi-view image, we select the salient feature points which the position and curvature change large to match the corresponding points of the model, in order to reduce the difficulty of feature extraction for automatic detection. A multi-view perspective texture mapping relationship is established to avoid the difficulty and complex computation of the one-time registration. Experiment of color texture reconstruction in tri-colored glazed pottery of the Tang Dynasty and Terra Cotta Warriors show that,2D-3D texture mapping method is satisfied with the color restoration model.The thesis is supported by National Natural Science Foundation Project "Research on virtual restoration of damaged Terracotta Army based on global optimization " and Shaanxi science and technology plan project of innovation project "Research and demonstration of the key technology of cultural heritage resources ".