在色彩領域裡很容易會遇到同色異譜的現象，這也是色彩複製上必須要解決的重要問題，然而透過多頻譜技術可以重建可見光的反射頻譜，換算出人類可感知的色彩三屬性明度、色相及彩度達到同色同譜，也能夠重建色彩與記錄色彩特性，這也就是色彩複製的最關鍵所在。
相較於以往的多頻譜掃描儀性能以及物理特性的不同，本研究採用分光式多頻譜掃描儀進行多種頻譜演算模式分析，從380nm~730nm內取樣間隔10nm的方式取出35個色頻波段，運用PCA與SVD期望能依其特性建立最佳演算分析模式，並且簡化其複雜的分析流程，以不同分析法特性之研究分析數據，頻譜的特性決定對色彩的判斷，所以在分析法上面的特徵分析更為重要。
本研究採用較先進分光式多頻譜技術對於瓷器進行研究，相較於以往多頻譜影像技術在於平面不同，以主成份分析與奇異值分解進行分析，並且從實驗上的結果發現，SVD在分析色彩的獨立性上較PCA佳，SVD除了分析青花瓷的主要成份較PCA穩定性外，從L*a*b*標準差可以看出SVD較能夠分析取得單色相之色彩主要成份。
The major task in color reproduction is to record the color characteristics and reconstruct the color correctly. However, the problem of metamerism is encountered frequently, which can be solved by multi-spectrum technique.
In this study, a grating-type multi-spectral camera was used to capture color signal from 380nm to 730nm in 10 nm interval. Further analysis was performed to calculate the spectral characteristics of the captured artifacts, mainly blue-and-white porcelains and simulated aged objects. Principal component analysis (PCA) and singular value decomposition (SVD) methods were used to analyze the spectral information into linear combination of the numerical color primaries (colorants). The efficiency of these two methods was compared by various experiments to find out which method is more suitable to current configuration. CIE Color difference values were calculated to provide the index for better performance.
The results indicated that SVD method performs better than PCA, and SVD method is more stability than the PCA method in reconstructing the color primary with less standard deviation in CIELAB units.
Keywords: Multi-Spectral Imaging, Principal Component Analysis, Singular Value
Decomposition
