[Openicc] Drop size calibration

[Graeme Gill]

Gerhard Fuernkranz wrote:

> But if we close down the first first drop size before reaching 100% area
> coverage, this implies that the next drop size should basically start
> dithering into the "holes" which have not yet been filled with the
> previous drop size, in order to obtain the least visible dot pattern.

That might be ideal from one perspective (smoothness), but not
so good for avoiding the 100% fill pattern. The approach I've
taken is to make the level index into the threshold array larger than
the range of two dot sizes. This means that the bigger drops that first
get set, are the ones that progress to the next biggest drop size first.
(See <http://www.argyllcms.com/dpssrc.html> for the implementation
of such screening), thus avoiding a 100% fill of one dot size at all
the intermediate levels.

> Wouldn't this - in the optimal case - require a threshold array for the
> 2nd level which is particularly tailored to the ink limit of the first
> level? Or do you think, it does not matter?

Yes, there are magic numbers that set the overlap level, although
it's usually pretty forgiving and works well over a variety of
inkjet engines.

> Isn't the "dot gain" (i.e. effective area coverage (computed from
> measured densities) minus prescribed area coverage (in %)) zero by
> definition, at 0% and 100% fill, and has its maximum rather somewhere in
> the middle?

Yes, I was really referring to the steepness of the curve as dot
gain starts to take effect, and then the saturation effect
once dot gain is dominant. Using a maximally dispersed dot stochastic
screen (good for smoothness), the light tints have practically no
dot gain, but as the dots start to be put next to each other,
dot gain starts to take effect. I'm not sure that you can say
that 100% coverage has no dot gain. Depending on the dot size
and placement, a 100% dot coverage with perfect dots may not
cover 100% of the area, whereas with dot gain added it may then
do so.


Graeme Gill.