So it sounds like just enlarging the print gives you larger dots, and
you can see them since there's no "data" between them...
Almost. The most basic method of enlargement, pixel doubling, replaces
each pixel with four (or nine, or 16) identical ones. That will only work
when you double (treble, quadruple, etc.) the image width and height, of
course. That's not a problem if all your image consists of is vertical and
horizontal lines but leads to jagged edges in all other cases.
To see the effect, create a new image in your editor. Make it a bitmap
(two colour) image with a white background and set the foreground colour
to black. Draw some straight, single-pixel lines at, say, 45 degrees and
about 60 degrees. Zoom in on the lines and see how they become very blocky
and jagged.
To get round the 'jaggies', as this effect is sometimes known, you could
fill part of each 'step' in the jagged line with pixels that are the same
colour as the line. If it's a 45 degree black line on a white background,
you could fill a triangular area of the step with black pixels, reducing
the step size and making the line appear smoother. If the line's not at 45
degrees it gets more complicated and some of the pixels have to be a shade
of grey to fool the eye into thinking there's a smooth edge. That's called
'anti-aliasing' and it's one of the most basic processes used in image
enlargement.
To see this effect, convert your bitmap image to greyscale, double or
treble its size and zoom in. If you include a vertical or horizontal line
in your bitmap image and increase the image size by a non-whole number,
you'll see that anti-aliasing is used on these lines as well.
If you're not increasing the image size by a whole number, it gets a lot
more complicated. You have to calculate new values for most of the pixels
in the image. When there are more than two colours in the image, and
especially when not all transitions between colours are hard-edged, it
gets even more complicated, because you have to calculate what colours to
use to make the colour transitions in all directions reflect the original
transitions as faithfully as possible.
...but Qimage, from what I've read in this thread, uses a 'triangular'
dot pattern when upscaling, which fills in the area much better.
I'm don't know how pyramid interpolation works but I'm pretty sure that it
doesn't refer to triangular patterns of pixels. It's probably more likely
to refer to the steps used in working out what colour each pixel has to be
and represents the 'tree' of decisions to be made, which either spread out
from the original calculation, getting closer and closer to the final
colour, or close in on it from a series of initial approximations. However
it's done, it's pretty smart stuff (well worth $45!) and produces very
good results.
Jon.