基于多光源维纳估计的光谱重建算法

Spectral Reconstruction Algorithm Based on Multi-Light Source Wiener Estimation

提出了一种基于多光源下彩色数码相机原始数据的物体反射率预测算法。通过测量获得彩色数码相机的灵敏度函数和每个光源的光谱,建立包括噪声项的相机原始数据与物体反射率的基本线性方程。通过训练色卡(X-Rite 140色色卡)估计系统噪声,并通过多光源维纳估计算法预测反射率。用两台不同品牌的专业数码相机对本算法进行评估。结果表明,本算法在D65、U30和HZ光源下的表现最优。相比分光光度计得到的光谱反射率,Canon相机和Nikon相机对孟塞尔色卡的预测精度较高,平均CIELAB色差分别为1.59和1.40,在非接触测色和颜色复制等领域具有重要应用价值。

This study proposes an algorithm for predicting objects’reflectance based on the color digital camera raw data under multilight source.The color digital camera’s sensitivity functions and each light source’s spectrum are measured,and the basic linear equation linking the camera raw data under multilight source and object reflectance,including the system noise term is developed.The training color card(XRite 140 color card)is employed to estimate the system noise,and the multilight source Wiener estimation algorithm is used to predict the reflectance.Two different brands of professional digital cameras are used to evaluate the algorithm.The findings demonstrate that the algorithm has the best performance under D65,U30,and HZ light sources.The Canon and Nikon cameras exhibit higher prediction accuracy for the Munsell color chart compared with the spectral reflectance obtained by the spectrophotometer,and the average CIELAB color difference is 1.59 and 1.40,respectively.It has crucial applications value in the fields of noncontact color measurement and color reproduction.