Abstract: A calibrating method based on tetrahedron interpolation is proposed. Because the sampling data is irregularly distributed in L*a*b* space, it is hard to construct 3D look-up-table for color conversion, which restricts the use of 3D interpolation in calibration algorithms. Consequently, an optimal vertex-finding algorithm in L*a*b* space is developed for constructing tetrahedrons, and the tetrahedron with smallest volume and enclosing the given point are selected, which is used to take the conversion from L*a*b* to device space. Experimental result shows the average error is less than regression and neural network method, which indicates that it can be used to calibrate most of the color devices.

Key words: calibration, tetrahedron interpolation, polynomial regression, neural network

摘要： 提出一种基于改进的四面体插值技术的校正方法。根据给定点将L*a*b*样本分区，通过比较体积找出包含给定点的体积最小的四面体，利用四面体插值及各顶点信息完成L*a*b*到设备空间的转换。实验过程中对打印机进行校正，采用多项式回归及神经网络等方法与该算法进行比较，结果显示该算法的平均误差小于其他2种算法，可以应用到彩色设备的校正中。

关键词: 标定, 四面体插值, 多项式回归, 神经网络

CLC Number: 

• TP 301.6