Re: Spikes in the reflected spectrum

Erik Nikkanen wrote:
> The reason I was asking was because if objects have relatively smooth
> reflected spectrums, that would imply to me that one does not need so many
> regions along the spectrum to give a reasonable description of the curve. 
> Maybe six or nine regions along the spectrum would be enough to represent
> and duplicate a curve to the point that it would look reasonably close.
>
> Erik

Tom Lianza and Danny Rich have interpreted your question as a matter of 
instrumentation and accuracy only. I went in another direction.

I am not the only person to represent spectral reflectances by a 
so-called "linear model." Usually when this is done, the goal is to gain 
some immediate advantage in the measurement lab. It is to understand the 
larger interactions that occur when people look at objects. Each eye has 
about 6.5 million cones, and there are a million fibers in each optical 
nerve. What stimulates the eye are black-white and color contrasts among 
objects.

Let's assume that good instruments are plentiful, a fact that Danny Rich 
has written into an ASTM standard. So you measure one object's spectral 
reflectance precisely at small wavelength steps. Then you measure the 
SPD of one light with similar precision. In a few microseconds my 
computer can calculate the tristimulus vector (X Y Z) from these data. 
And now what? One tristimulus vector tells you nothing. The same tsv 
could mean that you have a white object under a yellowish light, or a 
yellowish object under bluish-white light (say D65). Your eyes will use 
their millions of receptors to see the object in context and see the 
yellow object for what it is. In fact, the bluish light is excellent and 
desirable for seeing yellow objects, because they contrast better with 
white when there is ample blue for the yellow objects to absorb.

People who work with orthogonal basis functions, in one way or another, 
are usually concerned with larger issues of objects in context and/or 
the effects of the light.

My new articles differ from others that you might read for a couple reasons:
1. I am concerned with immediate practical implications. I work through 
a new formulation and apply it to data, seeking implications for 
everyday life.
2. Somewhat by luck I used color matching functions as basis functions. 
This leads to a convergence with Jozef Cohen's work.

None of this is to say that basis function methods cannot relate to 
instrumentation ideas. If an instrument computes an object's tristimulus 
vector under a standard light such as D65, that is consistent with my 
color-rendering model.