一种面向打印的低维光谱空间构造法

基于Kubelka-Munk混色理论模型,对其光谱吸收散射比空间进行线性修正,对光谱反射率空间进行正态化处理,并在两种空间之间建立经验变换,之后利用非线性优化技术和主成分分析方法,得到一个保持打印基色光谱特性的低维空间,将光谱反射率数据变换至该空间,以实现针对彩色打印的多光谱数据降维。实验结果表明,新方法能使低维空间数据保持高维光谱空间数据的最主要信息,克服传统主成分分析法导致重建光谱超出光谱数据范围的缺陷。和主成分分析法相比,新方法的色度精度、光谱精度以及同色异谱指数三个评价指标都有提高。

Construction of low-dimensional spectral space for multi-ink printing

By correcting coefficient ratio space of the absorption based on scattering of KubelkaMunk turbid media theory,and exerting a normalization transformation to the spectral reflectance space,a new linear space is created.Then a transformation between the spectral reflectance space and the new space is established.A low dimension spectral space which is consistent with the multi-ink color space is finally achieved by utilizing the nonlinear optimization and principal component analysis.The dimensionality reduction of the multi-spectral image for multi-ink printing is accomplished through the transformation of spectral reflectance to this low dimension space.Experiments show that the data obtained by this new method in the low dimension space can keep the main spectral information of the data in the high dimension spectral space.Furthermore,the new method can overcome the shortcoming brought by the principal component analysis,which is that the spectral reflectance reconstructed from the low dimension space may exceed the range of the spectral data.Compared with the principal component analysis method,this new method can make great improvements in the colorimetric accuracy,spectral accuracy and metamerism index.