Stefan,
Most of this is pretty subtle but ... you asked about accuracy.
--
On Mon, Aug 15, 2016 at 1:49 PM, Stefan Christiansen stefanchristiansen@... [QuadtoneRIP] <QuadtoneRIP@yahoogroups.com> wrote:
Thanks for answering unclear questions.
Not sure of what I want either, I ignore too much about linearizing with a spectro…
I guess the droplet needs real readings from a print.
I thought curves in PS with 16-bit images had a 16-bit accuracy, or at least were less destructive than with 8-bit images. I probably don't understand what accuracy means here.
PS curves are a bit of a hybrid. The points you enter are stored in 8-bit values but all the
interpolation when applied is done in 16-bit (assuming your image is 16-bit). So all the
intermediate points are nice and smooth. For the most part it's never an issue to worry about.
But for very flat or very steep curves like are sometimes used in QTR driver curves it can make
a difference. Let's take your example curve -- "0;0 50;75 100;100" -- in PS the 50 and 75 are stored
as 8-bit values. I.e. 127 and 191, which are actually 49.80% and 74.90%. If done in QTR driver curves
they would be 32767 and 49151 which are 49.999% and 74.999% -- that's the difference in 8-bit
versus 16-bit.
You also mentioned in your first post -- a 5% value of 720 vs 642 -- I assume these are .quad
numbers so 16-bit -- if they were only 8-bit they both lie between 2 and 3 (divide by 257) so mostly
indistinguishable.
But like I said at first this is pretty subtle differences. It would be hard to see it in prints.
I once checked that a "0;0 50;75 100;100" QTR curve and a .acv curve with similar values would create very close profiles (comparing the two profiles in QTR-CurveView or TextEdit). Then there are 2 possible L values for the 50% input in PS, and 3 possible L values for the 75% output…
So I assumed the maths behind both curves generators were the same, as well as accuracy.
QTR-CurveView offers a nice visual display of a profile's curve(s), especially in a full screen window (I already mentioned the coordinates not moving well along in an enlarged window). It's eloquent to compare two profiles with quick keyboard shortcuts switches.
It would be great if one could read x and y values (% or L) by pointing on the curve or moving a slider on one axis.
Would it be possible to make that happen some day?
There are no L-values in .quad's but just reading out the (x,y) percents sounds like a nice feature.
Roy
Thanks.
Stefan
------------------------------------
Posted by: Stefan Christiansen <stefanchristiansen@...>
------------------------------------