Finally, it was important to see just how well (or poorly!) the functions based on our measurments would predict the actual output of the camera. This was accomplished by essentially taking the original SPD values and processing them all the way through again.

The SPD's had the tungsten light removed from them (as presumably the camera would do.) They then were multiplied by the SPD2RGB matrix describing the color matching functions. This produced RGB values that then were "unbalanced" and had the gamma correction added back in. Thus all the camera processing was taken care of. This was done for all of the 113 samples that orginal data had been collected on.

It appeared that both the color matching functions from our math, and that of Professor Wandell's were only very general approximations of the camera CCD responses. Below are histogram plots showing the differences in RGB values between the estimated function and the actual measured RGB's of the camera response. The standard deviation here is approximately 50 for our function and 45 for the function generated by Professor Wandell's math.