Re: For those who are interested in making digital negatives using pigmented inksets

From: Bryan Lyles ^lt;[email protected]>
Date: 01/02/04-02:43:31 PM Z
Message-id: <>

At 06:30 PM 1/1/2004, wrote:

>I was looking for information on what the human eye can resolve and so far
>have come up with something in the neighborhood of 300 ppi—if anyone has
>information on that, plus thee number of gray levels that the human eye
>can distinguish (I've seen everything from 200 to 1000) I would appreciate
>the site.
>Mark Nelson

[Disclaimer: First of all, I am a computer scientist, not a xerographer, a
human factors person, or a photo-scientist. That having been said, I have
had a bit of experience with laser printers and imaging, having worked at
Xerox Webster in the 1980s and having hung around with some very good
graphics folk when I worked at PARC. Most of this information is from 10
to 20 year old memories Take what is written below with appropriate
amounts of salt. Real photo-scientists or xerographers may need to correct

When we were considering designs for the electronic imaging subsystems for
very high performance color printers (100+ pages per minute) I had the
library order a copy of Ed Grainger's PhD thesis (University of Rochester,
Institute of Optics, sometime in the 1970s) from University Microfilms. I
understand that Ed was one of, if not the, first to measure the modulation
transfer function of the human eye. He is also worked at Kodak and was the
source of the subject quality factor measurements that Pop Photo uses when
testing lenses. Translating Grainger's charts into pixels per inch at
normal reading distances, you get about 400 pixels per inch at 16 grey
levels per pixel as the limit of human vision.

This does not mean that the eye is not sensitive to higher resolutions in
specific cases. In particular the vernier acuity of the eye is much higher
and can detect mis-aligned lines at well over 1200 pixels per inch if they
are solid black. This is highly important for computer graphics but far
less so for photographs of natural objects. Nor does this mean that the
eye cannot see more than 4 bits per pixel for large smoothly shaded
areas. My memory is that the peak of the eye's sensitivity to the number
of grey levels is for spatial frequencies which translate into image
resolutions of around 100 pixels per inch. At that resolution/spatial
frequency the eye is many times more sensitive to contrast. Color accuracy
and the eye's ability to distinguish color differences at differing spatial
frequencies is distinct from the luminance MTF. Read Grainger's thesis or
more recent studies if you really care about the exact numbers. The
resolution needed and the number of bits per pixel is very much a function
of subject matter! A claim of "this picture looks good at x pixels/inch
and y bits/pixel" tells us very little except about the specific picture.

Now for a complete diversion:

For reflective prints, ie standard photographic prints, it is difficult to
get a dmax of greater than about 2. This limits the density range needed
for prints. Some of the 1980 era copiers had higher Dmaxes (3+) but this
was by building up heavy layers of toner and there was no reflections from
the underlying substrait. For transparencies the visible density range is
much greater. At any rate, these numbers and Grainger's results lead to
half-toning requirements for photo reproduction using digital
reprographics. Given a printer and its xerography, laser, modulator and
optics, you pick the halftone cell size to balance the spatial resolution
requirements and the grayscale requirements, ie if possible match the MTF
curves of the eye.

For older xerographic systems it was important to run the equipment so that
the process contrast was very high. This provided process margin since
setpoints could wander without black becoming grey. Furthermore, if you
are copying text people like really sharp (high contrast) results. As a
result the early systems could only write binary (white or black) onto the
photo-receptor. If you had a halftone cell of m by n pixels then you could
get m*n+1 grey levels (if you xerography was clean). This comes from 0
pixels within the halftone cell being black, one pixel black, ..., m*n
pixels black. The resolution of the halftone is then the base resolution
of the printer, for example 600 pixels/inch, divided by m or n (depending
on the direction).

Newer xerographic systems have greatly improved xerographic process
controls and can hold a process setpoint. Toner particle sizes have also
dramatically decreased and as a result it is easier to get uniform coverage
with less toner. As a result you can consider using multiple levels of
grey. Interestingly enough, you don't want to use simple fractions or
multiples when chosing grey levels. For example, if you had a 2x2 halftone
cell and can write the levels 0, .25, .5, .75 and 1 you would expect to
have 2*2*5 (or 20) distinguishable levels. You actually have far fewer
distinguishable levels since, for example 4 entries of .25 yield the same
result as one entry of 1. Among other considerations, this means that you
should pick the levels to be relatively prime (in the sense that none are
simple multiples of others). You also want to pick values so that they
result in an even set of density steps.

The story when using error diffusion or other stochastic halftoning methods
is somewhat, but not radically different. There is a fundamental tradeoff
between the spatial resolution, the number of grey levels provided by the
imaging system and the number of distinguishable levels in the print.

Why did I go into the long rant above? Because many of the results learned
in the digital xerography world should also be applicable to ink-jet
negatives, especially ones produced by inkjet printers with variable drop
sizes or multiple dilutions of the same color. They should also be true
for negatives produced via "colorization" where the different colors have
differing (probably relatively prime) transmission densities at UV
wavelengths. I have a strong suspicion that significant improvements in
the quality of our digital negatives are possible without changing the
basic printer hardware. For example, I could imagine a ink set which was
optimized for making colorized digital negatives when using a particular UV
light source (for gum, Pt, etc). Is there an photo-science student on the
list looking for an honors thesis?
Received on Fri Jan 2 14:41:57 2004

This archive was generated by hypermail 2.1.8 : 02/02/04-09:49:58 AM Z CST