guidelines for authors of cave surveys (Fri Oct 25 16:13:11 2002)

Andy Waddington on Survey stuff
Sat, 2 Nov 2002 19:19:07 BST

Wooks said:

> For scanned surveys ... a PNG or GIF will be something like 5%
> of the size of the scanned bit map

For surveys reproduced at a smaller scale, you tend to need thicker lines,
and GIFs and PNGs perform about the same. When reprodiced at higher
resolutions with thinner lines, PNGs perfomr significantly better in terms
of compressed size in my experience. However, a lot of that experience is
in terms of rendering vector images to bitmaps at printer resolution, rather
than actually scanning from an already rendered (ie. printed) survey. When I
do the latter, I tend to spend very many hours cleaning up pixels, 'cos I'm
a perfectionist :-)

However, back to what Wookey was saying:

> A 256-grey (8-bit) or higher scan just has a lot of fuzziness ...

I get the cleanest scans by scanning (or rendering from vector format) at
the very highest resolution that my scanner/amount of memory will permit
(600 dpi for huge surveys, 1200 dpi for stuff that can be fitted on a few
sheets of A4), then reducing the resolution to what I actually want later,
at the same time as making the pixel depth 2bpp grey. If I'm scanning from
clean copy then it is also usually best to scan at 1 bpp, and "anti-alias"
by transforming to lower resolution at the same time as moving to 2 bpp. You
can also fiddle around with the gamma to make lines come out thinner if you
scanned (or were sent a scan) at a high pixel depth.

However, note that the 2 bpp images are only useful if you have a rendering
device (screen or printer) which can render grey pixels ! If you are going
to print line art at maximum resolution on a laser printer, for example,
then it is better to keep more pixels at 1 bpp rather than doing this
anti-aliasing. It is precisely this sort of dependence on the eventual
output device that makes a vector format preferable for a computer-generated
(rather than scanned) graphic.

On several surveys which I was obliged to reproduce from scanned originals,
I found it vastly better to drop the bit map into a drawing package, and
carefully redraw the entire survey in vector format, which enabled me to
use exactly the line thicknesses that were appropriate and still get a
vastly smaller file. It only took between a few hours and a few days for
each survey, and *never needs to be done again*. This may sound laborious,
but it is a lot quicker than pixel-editing out all the unwanted lines
when you scan a survey drawn on graph paper :-)   

If you *are* doing vector format, then that has the advantage that it will
reproduce more and/or crisper detail on your printer's typesetting machine
than you will see on any home-computer printer. The sort of resolution you
need to do that with bitmaps really are going to be huge.

Praezis says:

> try to avoid A3 or A1 ...

We are talking digitised surveys here - software needs to abstract whatever
level of detail is required. With systems like KH, we are avoiding A0
because it is far too small to reproduce the detail recorded in the cave.
Trying to produce something that is meaningful when reproduced at A4 needs
not merely a vector versus bitmap discussion, but some sort of structured
picture which controls what details are reproduced at various scales, much
the way graphic games or scenery renderers for flight simulators don't try
to draw all the detail in distant parts of the view. This is probably a very
fruitful area for discussion as cave survey software becomes increasingly
sophisticated, though perhaps outside Dave's remit as an editor of an
A4-format rag :-)

However, at the end of the day, the best advice I can suggest you put into
your guidelines is "Do not submit anything, unless you are a dedicated
perfectionist with vast amounts of spare time to GET IT RIGHT!!!" as this
will save you having to do all the donkey work.

Of course, by the same token, as an author/survey artist, the only
acceptable deadline is "when it is _properly_ finished".