Juddj KacAdam, and Wyszecki^ have shovm that the 

 variations v.'hich occur in natural daylight (sunlight + skylight) 

 can be fitted with two characteristic vectors . The average 

 curve and the first two characteristic vectors for their data 

 are shov/n in Figure 5« The spectral energy distributions of 622 

 samples of daylight can be matched by linear combinations of these 

 curves . 



Characteristic vector analysis reduces the dimensionality 

 of a set of data to a minimum. Each of the daylight spectral 

 distributions can be represented by a point in a tv/o-dimensional 

 space. Each of the algae absorbance curves can be represented 

 by a point in a three-dimensional space. This is illustrated in 

 Figure 6. The coordinates of this three-dimensional space are 

 the vector amounts Y]_-y2-Yo. Three points are shov.-n plotted in 

 Figure 6. The locations of each of these points is determined 

 by hov; much of each of the vectors are required to match that 

 sample. 



The spectral sensitivities of a color film define another 



three-dimensional dpace. The coordinates of this three-space [ 



(Figure 7) are the red, green, and blue layer exposures of the 



31-9 



