It's been a long time since Ben and I agreed on anything.  This is one of
those times.  He's pretty much right on target.  He only left one thing out.
Learn to completely control buoyancy before you learn to use a camera
underwater.  Underwater photography is a significant task load.  Best to
reduce other tasks to automatic before you take that load on.

My older underwater camera is a Nikonos V equipped with a SB103 strobe.
It's not a slave, it's a wired strobe.  The combination worked quite well
once I learned how to use it effectively.  I quit using it because I was
spending more time taking pictures I wanted than I was enjoying the diving
itself.  I decided that people that wanted to see how wonderful the dive was
could take a course and do the dive for themselves.

Once again, I have a camera capable of underwater photography.   I have not
taken it underwater yet and mention it here only because it has features the
cameras Ben mentioned don't, at least as far as I know they don't.  I have
an Olympus Stylus 1030 SW.   It's a small 10.1 mp camera that is, without a
housing, good to 30 feet.  Olympus makes a housing and a slave strobe setup
for it that allows through the lens metering.  I chose it over other brands
(I have high end Nikon and Minolta land camera equipment) because, being
waterproof, it's also rain and moisture proof.  For someone still interested
in underwater photography, being waterproof increases the odds that it will
survive should the housing flood.  Non waterproof cameras rarely survive
even a partial housing failure in salt water.  The Olympus has an underwater
mode.

If you didn't look at Dan's photos, you should.  Some of them are quite nice
and most, if not all, were taken without a flash.  This works well for Dan
at least partly because he tends to dive shallow, where blue shift is not as
severe, and because he's a fair hand with Photoshop or similar photographic
software.  What's probably more important about Dan's photographs is that he
does not usually sacrifice his enjoyment of the dive to get them.

One of several things that bothers us older style photographers is that
photography skills these days include significant computer enhancement
skills.  Of course I don't talk about the enhancements I used to do with the
darkroom equipment that's still up in my attic.

Lee

Backscatter is a physical reality when there is particulate in the
water.  The only ways to minimize it are to:

a) Use strobes that are physically displaced away from the lens. See:
<http://www.huntzinger.com/photo/2007/backscatter.jpg>

b) Not use any strobes...shoot just ambient light

c) Not dive where there's particulates in the water...or at least, not
take any photos.

A 'Slave' isn't necessarily a must - it is the capability to have the
strobe head physically located away from the lens, which is different
than how the strobe is connected to the camera (wired/wireless) or how
it is controlled (manual/TTL or Slave).

In general, the best image will result from a hardwire connection that
is directly controlled by the camera and the worst by a wireless
slave.  The basic reason is that a slave is invariably less responsive
in receiving information for when to quench, which is necessary to
prevent an overexposure.

Two heads if you don't want deep shadows

The definition of "reasonable" depends on your expectations and your
budget.  I'm looking at $6K to replace my old film Nikonos V system (w/
dual strobes)

And likely more particulates, which accentuates the problems of
backscatter, as well as making for darker ambient conditions, which
will tend to make all autofocus systems take longer to focus.

There's two fallacies here.

The first is the paradigm of getting blood from a stone.  The red
spectrum is strongly absorbed by water, and while there are ways to
compensate by boosting the red channel ("Mandrake" technique), the
stark realities are that there are scenes where there simply is no
substitute for a strobe because the ambient lighting has become
monochromatic.  Photographic examples that I'd **love** that some
Photoshop post-processing trick would work miracles are available upon
request.

The second one is that what any one person finds acceptable is also
acceptable to others.  Here, I'm specifically referring to noise
issues at high ISO.  My current dSLR has equal-to-slightly-better high-
ISO performance than the F30 (per dpreview), yet I wish it were
better.  As such, I generally avoid using ISO 1600 except in those
circumstances where I can't avoid it, which typically are off a tripod
with long shutter speeds due to negative Exposure Values (EV's).

In general, there's limitations of physics at work and while the image
can be 'de-noised' through subsequent processing (some of which occurs
in the camera), there's an inevitable trade-off in image resolution
potential.  Ultimately, when these 'other factors' are made equal, it
is the size (area) of the digital sensor that is the primary factor in
how much noise is generated when the ISO gain is increased, and there
is no substitute for using "bigger film" (a larger sensor).  I've not
kept up on the new Nikon, but previously, the best high-ISO
performance in a "within reach of the consumer" camera came from the
Canon 5D, which features a full frame (24mm x 36mm) sensor, in lieu of
the DX sensors in most consumer dSLRs or the even smaller sensors that
are in the P&S's.  Thus said, I'd not exactly recommend the 5D today,
as it dates from 2005 and its replacement is immanent.

-hh

I have what's a fairly similar digital P&S with housing & diffuser and
in my experience, its performance is "surprisingly good" under benign
conditions, which I define as daylight tropical, but when the lighting
drops or the particulates increase (or both), the images become
snowstorms of backscatter.

-hh

"Just pick up any SCUBA rag from about 20 years ago and you'll see the
same
pics, most better, taken a thousand times by each and every Bonaire
diver
with a camera and little talent."