[Openicc] L* was: MS on HDR CMS :)

[Graeme Gill]

Chris Murphy wrote:

> Who cares if it matches human perception? That's not a requirement  
> for image encoding. Image encoding - RGB values are *device values*.  
> They are *device dependent*. They are merely a means of communicating  
> signals to *devices*. Who cares what the necessary tone  mapping is  
> to get them to do what we want? So long as they are consistent and  
> predictable and correlate through a correctly built ICC profile those  
> values are referenced back to LAB and to L* already.

It's not that hard. If L* really is perceptually uniform, then
it's also the ideal encoding space to ensure that quantization
errors are perfectly equally dispersed, hence optimally minimized.
So IF L* is visually more uniform than the sRGB gamma,
it would in fact be better if the native behaviour of
the device was L*, even if then other gamma's are emulated,
since L* would give the optimum control over the display
with a minimum number of bits. (I say IF, because L* is only
perfectly perceptually linear under a particular set of
viewing condition, and typical display viewing conditions
may not match. On the other hand, the straight line segment
in the sRGB response is a mathematical convenience that
may be compromising the visual uniformity.)

Graeme Gill.