QTR-Quadtone RIP

--- In QuadtoneRIP@yahoogroups.com, Roy Harrington <roy@...> wrote:

Thanks Roy for explaining this further to us. I realize now the 
linearization bit is that what I was implicitly assuming.


Joost

> 
> What Joost is suggesting is a bit simpler than I think you are 
assuming.
> We already have blending between two curves.   This gives you any 
hue
> along a straight line between the hue of each curve.   It's all one 
> dimensional,
> you just weight the two curves for every point along the L axis.   
> Using three
> curves is just a simple extension of that.  But it's still a 
> one-dimension weighting.
> With this scheme the possible hues are greatly increased because 
they 
> are
> no longer just a straight but a triangle formed by the points of 
each 
> curve.
> 
> In general this might not seem possible and probably wouldn't work, 
but 
> with
> QTR profiles this works quite well.   The first key issue is that 
the 
> curves that
> you blend are all linearized the same way.  Dmin and Dmax in 
general are
> identical and all the L values in between are the same for the same 
> grayscale
> value.  This makes the weightings always between the same density.  
The 
> other
> key issue is that the weightings are calculated deep in the driver 
> where the
> exact amount of ink is known.   These two facts are what make this 
all 
> possible.
> It's probably not 100% guaranteed that the blendings will always 
give 
> identical
> densities but in the tests I've done with two curves it's shown 
very 
> good
> consistency.  I expect 3 curves wouldn't be appreciably different --
 
> we'll see.
> 
> Note that this scheme is not likely to work at higher levels like 
in 
> ICC profiles or
> for the Epson driver in color or ABW mode, because there's no 
access 
> and/or
> knowledge of how much ink is actually being used.
> 
> Roy
> 
> -
> Roy Harrington
> roy@...
> Black & White Photo Gallery
> http://www.harrington.com
>