Treffer: 基于聚类算法的纺织品文物色彩提取与纹样数字化探索———以新疆巴里坤M12 出土清代纺织品纹样为例.

With the advent of digital simulation technology there is a new solution to the problem of secondary damage caused by the display and handling of textile artifacts which due to their special material properties are prone to suffer secondary damage in the course of handling for exhibition. In the process of digitally simulating the restoration of textile artifacts the colour extraction step is very important and this paper proposes to use computer technology to do this. The aim is to obtain accurate and efficient colors for textile artifacts without secondary damage. In order to accurately and effectively extract the colors of textile artifacts and to digitize the patterns of textile artifacts we taking a group of Qing Dynasty textiles excavated in Balikun Xinjiang as an example designed a textile pattern recovery method based on the multivariate fuzzy C-mean MFCM algorithm by combining with the K-Means + + algorithm . First images of textile artifacts were acquired and calibrated and the acquired images were smoothed and noise-reduced with a bilateral filtering algorithm. The MFCM algorithm was then used to segment the smoothed and noise-reduced image. The aim of this step was to remove areas of the image that were abruptly colored and areas where the artifacts were missing so that subsequent algorithms could extract the colors more accurately. In the comparison of the effectiveness of the segmentation results with the fuzzy C-mean clustering and enhanced fuzzy C-mean clustering algorithms the results of the MFCM algorithm were found to be better and the IOU evaluation value of the MFCM algorithm was the highest among the three algorithms indicating that the MFCM algorithm segmented the images with more accurate elimination of discolored and damaged parts of the images and less useful details of the images were lost. The K-Means + + clustering algorithm was then used to analyze the number of clusters and to select Buzi as the case. The above experimental results were compared and analyzed with those of the software colour extraction method and the K-Means clustering algorithm for colour extraction in terms of RGB colour scatter plot algorithm duration and RGB histogram. The colour scatter plot of the K-Means + + clustering algorithm was found to be closer to the RGB colour scatter plot of the artifact image and the K-Means + + clustering algorithm was found to be faster in related experiments and the RGB histogram results were closer to the RGB histogram of the artifact image. It shows that the K-Means + + clustering algorithm can effectively extract the main textile colors while well retaining image details and its experiments are also faster. Finally the pattern outline was extracted by computer software and the extracted main colors were filled into the pattern outline to complete the digital simulation of the textile heritage pattern which is also an important part of the virtual display of textile artifacts. The algorithm in this paper also visualizes the colour matching pattern of textile artifacts and the colour matching ideas of the sample artifacts can be seen through the hue distribution model and the lightness and saturation distribution tables. The hue distribution of the Buzi is on the cooler side while that of the python robe transitions up and down from warm to cooler tones and the heritage motifs are moderately saturated and of low brightness giving a solemn and sedate visual impression. This paper uses images of cultural relics physically photographed in Xinjiang as a research sample and expands on the clustering algorithm colour extraction to make the results of colour extraction of textile relics more accurate. After a comparative analysis with other algorithms it can be seen that the method proposed in this paper achieves an optimized effect on the colour extraction of textile artifacts. The experimental results show that this research method can extract the colour of textile artifacts more accurately and the extraction effect is better than the traditional digital colour extraction method and the colour extraction efficiency is higher. In addition after digitally extracting the colors of the motifs of the samples studied in this paper using computer algorithms we can visualize the colour scheme characteristics of the group of cultural relics which abstractly represents the colour style and cultural connotations of the motifs. The experiments designed in this paper have value for further research and application and can provide reference for conceptual solutions for the use of garment digitization technology to display virtual cultural objects facilitate the exhibition and dissemination of culture in the same category and gain new experience for the digital conservation of textile objects in the future. [ABSTRACT FROM AUTHOR]

为准确有效地提取纺织品文物的色彩,实现纺织品文物的纹样数字化处理,文章以新疆巴里坤出土的一组清 代纺织品为例,设计了一种基于多变量模糊C 均值(Multivariate Fuzzy C-mean, MFCM) 聚类算法与K-means + + 算 法相结合的纺织品纹样复原方法。首先,获取图像并通过双边滤波与高斯滤波对图像进行平滑降噪处理;然后采用 MFCM 聚类算法对平滑降噪后的图像进行分割;接着采用K-means + + 算法分析聚类数量,有效提取纺织品主色,并 较好保留图像细节;最后通过计算机软件提取文物纹样轮廓,将提取的主色填充至纹样轮廓,完成纺织品文物纹样的 数字化模拟。实验结果表明,该方法可以较为准确地提取纺织品文物色彩,并且提取效果优于传统数字化取色方法, 取色效率更高,具有进一步研究应用的价值。 [ABSTRACT FROM AUTHOR]