[net.graphics] resolution questions

tim (01/26/83)

At what resolution do you start to get really good print quality?

CRT's with 512 X 512 resolution still can not draw a straight line
from one corner to the diagonal.  1024 X 1024 seems to do okay (although
the only one I've ever seen has been the Tek storage tube displays).

In terms of dots/inch on hardcopy, 200 dots/inch on our Benson/Varian
plotter looks nice but vtroffing out the documentation for "Mathematical
Typesetting" really strains the resolution to a point that it begins
to look bad.  I seem to remember a disscussion on the net sometime back
that talked about 500 dots/inch being acceptable and 1000 dots/inch as
being printer quality.  Comments?

And my last point/question is:  Would a hard copy stroke graphic device
be useable vs. a raster device?  You still need to provide programmable
fonts and different point sizes etc.  Is this still possible and practical
with stroke (vector) drawing vs. raster (dots/inch)?

			Tim Curry
			USENET:  ucf-cs!tim
			ARPANET: tim.ucf-cs@udel-relay

ken (01/27/83)

    The question was at what point one can get good print quality.  I
take it this means fonts, although lines were mentioned in the original
article.  I don't know too much about printers, but I have had a lot of
experience with video.
    What is your definition of quality? How do you measure it? What
sort of things are most disturbing to you about print at 512x512?
    The things most disturbing to me are the ragged/jagged edges,
otherwise known as aliasing (in the sampling theorem sense).  Certainly
increasing the resolution to 4Kx4K reduces the jaggedness, but does not
eliminate it completely.
    To eliminate the aliasing (jaggedness) completely, one must resort
to anti-aliasing techniques (aptly named!).  When you draw lines,
circles, or splines, you are sampling a continuous mathematical
function at regularly spaced points on a rectangular grid, so sampling
theory must be taken into account.  Sampling theory says that you damn
well better low-pass-filter your signal before you sample it, lest you
get aliasing.
    Aliasing occurs because you round the location of an edge to the
nearest pixel.  One can move the apparent location of an edge between
pixels by diminishing the intensity:  Anti-aliasing techniques trade
off amplitude (brightness) for phase (location), resulting in a shift
in the apparent position of an edge.
    I used antialiasing techniques to render fonts, lines, and polygons
when I was a member of the engineering staff of a broadcast quality
video art system (AVA) a few years ago.  The frame store resolution was
768x486x8, and we got great looking characters.  The quality was so
good, that even characters *3 pixels high* were readable, as long as
they were upper case and well separated.  6 pixel characters were good,
and 10 pixel characters were excellent, especially compared with 3.
    The more bits per pixel that one has, the more one can get rid of
the aliasing.  The biggest jump in quality was from 1 to 2 bits per
pixel, although we generally used 4 (for intensity; the other 4 were
used for tint).
    Anti-aliasing tends to smooth out edges at the expense of making
edges look softer.  Increasing the resolution will make edges sharper;
3 pixel high characters will still be as readable as at lower
resolutions.

				Ken Turkowski (turtlevax!ken)

[1] Crow, F.C. A comparison of antialiasing techniques. IEEE Comput.
    Gr.  Appl.  v.1 no.1 (Jan. 1981) pp. 40-48
[2] Turkowski, K. Anti-Aliasing through the use of coordinate transformations.
    ACM Tr. Graphics, v.1 no.3 (July 1982) pp.213-234

bryan (01/28/83)

Last year the Seybold Report on Publishing Systems had a two part article
on digital type.  It is an outstanding survey of the problems encountered
when attempting to use low and medium resolution printers (less than 5 or
6 hundred spots per inch).  The figure given for photo typesetter quality
output was 2000 SPI.  This was on the basis of the reproduction of 4 pt
type.

I strongly recomend reading the articles if you can find them.  I have the
reference packed away but will attempt to find it if anyone needs it.

-Bryan Lyles
(allegra or seismo)!rochester!bryan

djb (01/31/83)

I've seen anti-aliased lines drawn on a 640x480 display and they
looked pretty good.  This technique has been applied to text displays
with similar success. Unfortunately, you must have a large color look-up 
table and a broad color palette to be able to produce the subtle shades
required for anti-aliased displays (graphics or text).

As for hardcopy, take a look at some of the 300 dot/inch laserprinters.
They do an excellent job with text and graphics, although clearly not as 
sharp as a typesetter.  We currently use a Versatec for document prepration
and are reasonably happy with it (you're right though, on some output it does
look coarse), and expect our soon-to-arrive laserprinter to be more than
enough for our needs.

Gee Whiz department:  I saw the Pixel graphics display terminal at UNICOM.
It has a 8.5"x11" white phosphor display with a resolution of 1728x2180.
The text and lines they drew were something else...

	David Bryant   Bell Labs   Columbus, OH   (614) 860-4516
	(cbosg!djb)