摘要：
Under given illumination and observation conditions, in order to reconstruct the spectral reflectance of the object surface from the camera response signal to achieve high-precision color reproduction, multi-spectral imaging technology was used to acquire multi-spectral image response of an object. The principal component analysis, matrix R and the new regularization matrix R method were used to analyze the theoretical analysis of spectral reflectance reconstruction. The results of the research were verified experimentally in the color reproduction of mural color blocks. The reconstructed spectral reflectance and color reproduction data of mural color blocks were obtained. At the same time, the color reproduction results of mural color blocks based on regularization matrix R method were evaluated. The results show that the regularization matrix R method is superior to the principal component analysis and matrix R method in the spectral accuracy and reconstructed accuracy. Compared with the principal component analysis and the matrix R method, the color difference is reduced by 0.0732, the fitness coefficient is increased by 1.10％, the root mean square error is reduced by 0.0035, and the spectral matching partial index is reduced by 0.0225. This method can meet the needs of high-precision color reproduction, which is suitable for digital archiving of cultural relic artwork, restoration of cultural relics and other fields.%为了在给定的照明和观察条件下, 用相机响应信号重建物体表面光谱反射率, 实现颜色的高精度复原, 采用了多光谱成像技术采集物体表面的多光谱图像, 使用主成分分析、R矩阵和正则化R矩阵方法进行了光谱反射率重建的理论研究, 并对壁画色块颜色复原进行了实验验证, 取得了壁画色块的重建光谱和颜色复原数据, 同时对基于正则化R矩阵方法的壁画色块颜色复原结果进行了评价.结果表明, 正则化R矩阵方法进行光谱重建的光谱精度和色度精度更高, 与主成分分析和R矩阵方法相比, 色差降低了0.0732, 适应度系数提高了1.10％, 均方根误差降低了0.0035, 光谱匹配偏指数降低了0.0225.该方法能够满足高精度颜色再现的需要, 适用于文物艺术品数字化存档、文物艺术品修复等领域.
