透過您的圖書館登入
IP:3.144.154.208
  • 學位論文

利用模式化色域極點同色異譜重建物體頻譜反射率

Recovering Spectral Reflectance Based on Interpolation Method with Model-based Metameric Spectra of Extreme Points

指導教授 : 周遵儒

摘要


印刷科技與媒體影像傳播等大量使用色彩作為媒介的產業,皆依靠彩色複製(Color Reproduction)這項技術進行色彩再現,這項技術隨著科技的進步不停的更新、改變,現今由於傳播媒材越來越多樣化,跨媒材的色彩管理顯得十分重要。多頻譜影像(Multispectral Image)能完整儲存的物體頻譜資訊,提供不同條件下的色彩模擬,是彩色複製未來的趨勢,然而頻譜反射率不易取得,因此研究者們試圖以三刺激值重建頻譜反射率。 本研究利用內插法重建物體頻譜反射率,並設計一組模式化色域極點同色異譜(Model-based Metameric Spectra of Extreme Points),來解決內插重建法在訓練樣本不足時的計算問題,將內插重建法的計算範圍擴大至目前數位影像常用的sRGB全色域,並以RMSE、GFC及色差公式評估物體頻譜反射率重建結果、色彩模擬結果。 本研究一共進行三組實驗,重建色票樣本、重建sRGB全色域、重建多頻譜影像。研究結果顯示MMSEP能有效的擴大內插法的計算範圍至sRGB全色域,MMSEP+LI法、MMSEP+NNI法(統稱MMSEP內插法)在頻譜曲線形狀重建上比PCA法更符合目標頻譜,MMSEP內插法在更換光源時色差的增幅也較小,但在數位影像轉換多頻譜影像上,MMSEP內插法重建結果並不好,頻譜曲線形狀有所差異,甚至在部分像素上出現較大的色差值。

並列摘要


The industry such as printing technology, image communication and other exten-sively use of color as media are relying on color reproduction to reproduce color. This technique keep updating and changing with the progress of technology. Nowadays, due to the diversification of media, cross-media color management become more and more important. Multispectral images can fully recode spectral information of the ob-ject and provide color appearance under different conditions, is the future of color reproduction. However, spectra reflectance is not easy available, so researchers are trying to reconstruct spectral reflectance with tristimulus values. In this study, the spectral reflectance of the object was reconstructed by using the sRGB values which were commonly used in the current digital images. A set of mod-el-based metameric spectra of extreme points was designed to solve the problem of the interpolation reconstruction method, the calculation of interpolation method was extended to sRGB color space. The reconstructed spectral reflectance of the object were evaluated by RMSE, GFC and color difference. There were three experiments in this study, reconstructed spectra reflectance of color chips, reconstructed whole sRGB color space, reconstructed multispectral imag-es. The results showed that MMSEP can effectively expand the calculation range of interpolation method to whole sRGB color space, MMSEP + LI method and MMSEP + NNI method (collectively, MMSEP interpolation method) more conform to the tar-get spectrum than PCA method in the curve shape of spectral reflectance. The MMSEP interpolation method showed a small increase in color difference when the light source was replaced. However, in transforming digital images into multispectral images, the reconstructed results of MMSEP interpolation method were poor. The shapes of spectral curve were different, and had large color difference on some pixels.

參考文獻


Abed, F. M., Amirshahi, S. H., & Abed, M. R. M. (2009). Reconstruction of reflectance data using an interpolation technique. Journal of the Optical Society of America A, 26(3), 614-624.
Agahian, F., Amirshahi, S. A., & Amirshahi, S. H. (2008). Reconstruction of Reflectance Spectra Using Weighted Principal Component Analysis. Color research and application, 33(5), 360-371. doi: 10.1002/col.20431
Amidror, I. (2002). Scattered data interpolation methods for electronic imaging systems: a survey. Journal of electronic imaging, 11(2), 157-176. doi: 10.1117/1.1455013
Ansari, K., Amirshahi, S., & Moradian, S. (2006). Recovery of reflectance spectra from CIE tristimulus values using a progressive database selection technique. Coloration technology, 122(3), 128-134. doi: 10.1111/j.1478-4408.2006.00019.x
Ayala, F., Echávarri, J. F., Renet, P., & Negueruela, A. I. (2006). Use of three tristimulus values from surface reflectance spectra to calculate the principal components for reconstructing these spectra by using only three eigenvectors. Journal of the Optical Society of America A, 23(8), 2020-2026.

延伸閱讀