中文摘要：

为了提高光谱分解效率,提出一种基于光谱方向的2DPCA的光谱分解方法,该方法可以直接利用原始图像矩阵进行光谱分解。通过对兔子动脉和多层油漆中红外显微图像的分解实验,验证了2DPCA光谱分解方法的可行性和有效性。实验表明,当光谱数大于通道数时,采用2DPCA的光谱分解效率高于PCA的光谱分解效率,并且随着光谱数与通道数比值的增大,2DPCA的优势更明显。

英文摘要：

In order to improve the efficiency of spectral unmixing,a novel spectral unmixing method based on 2DPCA in the spectral direction was developed.With the proposed method,spectra can be unmixed directly using the original image matrices.Simulation experiments performed on FTIR microscopic images of a rabbit artery and a multilayer paint chip verify the feasibility and effectiveness of the proposed algorithm.Experimental results demonstrate that when the FTIR microscopic image contains more spectra than spectral channels,the proposed algorithm has higher computation effi-ciency than PCA algorithm for spectral unmixing.Further more,the advantage of using 2DPCA becomes more obvious with an increase in the ratio of the number of spectra to the number of spectral channels.