The 'real' meaning of blend level transparency integers

[Jim Ault]

On 11/9/07 8:47 AM, "David Glasgow" <david at dvglasgow.wanadoo.co.uk> wrote:

> Hello folks,
> 
> I am writing a psychological assessment which involves visual signal
> detection.  Basically, a 'blob' slowly appears at a random point
> within one of three invisible rectangles.  I have just about cracked
> the random generation of a loc within a given object.
> 
> What I am having trouble with is understanding exactly what the blend
> level integers represent.  I am not image savvy, and don't know if
> these are, or can be related to, any recognised and objective
> transparency units.  Any recommended reading on this sort of thing?
> 
> Two specific issues are
> 
> 1/  A blendlevel set at 50 looks less than 50% transparent to me, and
> an incrementally decreasing transparency doesn't look linear to me.
> Is that to do with the technology or human visual perception?
> 
> 2/  How confident can I be that the visual effect of incrementally
> decreasing transparency will be equivalent across machines and OSs?
> I understand this is comparing Win 2000 and later with OS X.  Is
> there anything I need to watch out for?

One of your inconsistencies between computers and users is the monitor
settings (gamma, temperature, calibration) (cathode ray tube, low-cost LCD,
old cathode ray tube, expensive LCD)
Most users leave them set the same as 'out of the box' and graphic artists
calibrate to different color spaces, regardless of OS.  Of course, there are
parts of some operating systems that try to optimize the monitor for the
benefit of the user.  (for details about color space settings read the
Photoshop pdf manual on the science of setting monitors)

Another factor is the observer.  Some see colors and contrast differently,
and then others don't have good close vision (and out-of-date correction).
I learned this in an office where some users could not see certain colors
well, so we had to find a good font/background color combinations on some
database layouts.

As far as the Rev blendlevel, I would be surprised if it had any scientific
color space, gamut, or LAB calculations that related to human perception.  I
have only seen transparency calculated as a linear distance between 0 and
255, as in Photoshop.  This is to say, the calculation is linear, not the
human perception.

You probably need to 1) calibrate the monitors on some computers running
different OSes/monitors, 2) set the transparency+background of a test
pattern (like they do in TV broadcasting), 3) then build a curve that gets
you the 'human linear' transparency values.  Do one according to your eye,
then see if others agree with you.

I know there is a significant body of work out there for the printing
industry, but you may have to dig to find anything for your particular
effect.

my 2 cents

Jim Ault
Las Vegas