From: "Bob May" 
To: "atmlist" 
Reply-To: "Bob May" 


I've been using various security cameras over the last 4 years for testing
and have found that the first thing in any system is to insure that the
camera doesn't change its gain or any other variable as you do your
testing. With my present system, the camera works just fine with uncoated
mirrors. Coat the mirror and the testing becomes a lot more difficult as
the aluminum coating reflects so much light that the camera starts AGCing
and that means that the brightness of a zone becomes a lot more difficult
to determine and there is a large amount of residual brightness left after
the zone is dimmed that makes seeing the shadows a bit more difficult.
After that, a zoom lens is nice if you are doing many different focal
ratios but that is a luxury as a single lens will handle a supprisingly
large range of focal ratios without any real problems. With a manual zoom
lens and focus, any camera can pretty much do the job, even the little
320x240 cheap security cameras - an 8" mirror means that it is
0.025" per pixel when the mirror is the full width of the display and
still less than 1/16" wide on the mirror when the mirror is only
taking up 1/2 of the display width.
I'd recommend against the color cameras as they use a matrix system to
obtain the color and that means that the actual resolution can be less than
optimal but the real thing here is that the image will be majorially only
in one channel of the color and one thing about the color system is that it
is really low resolution compared to the luminance info.  In other words,
you think you're getting 640 pixels width but you are actually getting only
320 or even less.
As to the JPG info loss, Yes, there is loss of info in a JPG image.  The
loss starts with the loss of sharpness of sharp edges (go look at text that
has been compressed with JPG for a start) and at high compression ratios
goes more towards loss of the actual brightness levels (an 800K image
compressed to 10K or less in size for example).  For the moderate to light
compression ratios used by most, the actual loss of the image is often on
the order of a dozen or so pixels that have actually had their values
changed by a count or two.  This isn't what I'd call a major loss of info
that many have blowen it up to be.  What makes it more irrevelant to what
we are doing is that the major loss of the data at those levels is at the
sharp edges where there is diffraction problems anyway. I say these things
on the JPG  compression standard as I was part of the crew that developed
it and we did all kinds of testing to determine what actually was happening
to the images after numerous compression/decompression cycles and so had
the software programs to analyze the changes in an image.  The first
compression cycle usually took all of the losses while later cycles didn't
change anything.  Increasing compression always took more info out of the
image.  We used both "busy" images and "unbusy" images
(one image was a slow shift of color and brightness across the image kind
of like the shading of the sky from zenith to horizon) to verify the
operation of the compression mechanism in all variations of images.
In other words, don't be afraid of JPG as it doesn't lose that much info. Bob May
http://nav.to/bobmay
bobmay{at}nethere.com
NEW! http://bobmay.astronomy.net

--- BBBS/NT v4.01 Flag-4