影像科学与光化学 ›› 2005, Vol. 23 ›› Issue (5): 340-350.DOI: 10.7517/j.issn.1674-0475.2005.05.340

• 研究论文 • 上一篇    下一篇

用改进的基向量法重建光谱反射比

王葛1,2, 李长军3, 朱云龙1, M Ronnier Luo3   

  1. 1. 中国科学院, 沈阳自动化研究所, 辽宁, 沈阳, 110016;
    2. 中国科学院, 研究生院, 北京, 100039;
    2. Department of Colour Chemistry, University of Leeds, Leeds LS29JT, UK
  • 收稿日期:2005-04-01 修回日期:2005-06-01 出版日期:2005-09-23 发布日期:2005-09-23
  • 通讯作者: 李长军

Improvement in the Estimation of Reflectance Functions Generated using the Basis Vectors

WANG Ge1,2, LI Chang-jun3, ZHU Yun-long1, M Ronnier Luo3   

  1. 1. Shenyang Institute of Automation, Chinese Academy of Sciences, Shenyang 100016, Liaoning P.R. China;
    2. Graduate School of the Chinese Academy of Sciences, Beijing 10039;
    3. Department of Colour Chemistry, University of Leeds, Leeds LS29JT, UK
  • Received:2005-04-01 Revised:2005-06-01 Online:2005-09-23 Published:2005-09-23

摘要: 如何从已知物体表面颜色的三刺激值及其照明条件和观察条件准确地重建其光谱反射比,是彩色图像研究领域尚待解决的一个重要课题.本文首先介绍了两种前人提出的方法,即“伪逆矩阵法”和“Wiener方法”,然后,分析了这两种方法的优缺点,指出了共同存在的问题,即所重建的光谱反射系数有可能超出实际可能的范围(0,1),进而提出了一种新的改进的基向量法.最后,文中还给出此改进的算法和上述两种算法的特性模拟仿真.从仿真的结果中可以看出,改进的算法不仅在精度上较上述两种算法更精确,而且能够保证所重建的光谱反射比能够满足其实际可能的范围(0,1).因而,用本文中所给出的改进的基向量法重建物体表面的光谱反射比,较上述两种前人提出的算法更能满足工业实际需要.

关键词: 光谱反射比, 奇异值分解, 基向量法, 伪逆矩阵法, Wiener法

Abstract: Various techniques based upon basis vectors have been developed to generate reflectance functions from a given set of tristimulus values or an input devices’ RGB responses. At first two widely used methods, the Generalised Inverse and the Wiener, were introduced. A drawback of these methods was addressed, which is that the generated reflectance functions are sometimes out of the range between 0 and 1. A new method was developed to overcome this problem. This method was tested together with the other two methods. The new method not only gave more accurate predictions to the test data but also satisfied the boundary conditions. Furthermore, it was found that using more than 5 basis vectors does not perform better than those using 5 basis vectors.

Key words: reflectance function, tristimulus values, singular value decomposition, basis vectors, Generalised Inverse method, Wiener method

中图分类号: