[OT] Digital camara

Les hlhowell at pacbell.net
Tue Nov 6 15:55:47 UTC 2007


On Tue, 2007-11-06 at 13:17 +1030, Tim wrote:
> On Tue, 2007-11-06 at 10:53 +0900, John Summerfield wrote:
> > The more pixels does translate to better printed images, and my local 
> > camera vendor assured me 3.2 Mp is good to about A4. Which probably 
> > translates to marginal to the trained eye. 
> 
> I found my 4 meg camera to just about manage an A4 sized print, but only
> for distance viewing, and only when the source photo was taken under
> good conditions.  Close up, you could see JPEG artifacts, and artifacts
> from the camera and printer (I use the local photo mini lab for my
> prints) that try to sharpen the picture more than it really is.
> Unfortunately they nearly all try to crispen pictures, rather than print
> pictures as they are, or use a camera which has better native resolution
> (that's optical and electronic, in combination), in the first place.
> 
> Nearly all electronic imaging relies on artificially peaking the
> resolution.  Turn off the detail enhancement, and you think "yuck, it's
> very soft," like the old soft filters used on female movie stars, even
> on $20,000 cameras.  Conversely, for film, back when you could still get
> your photos processed optically, and you had a decent camera, even a
> moderately priced consumer camera, we had better looking prints (from
> the decent labs).
> 
> Granted that enlargements tend to be viewed at a distance, and that does
> hide some defects, but some are not viewed that way.  Such as family
> photos sitting in a frame on the desk.  They'll probably not be A4
> sized, but still can be big enough to expose some failings in the
> system.  Never mind someone who wants to be arty, and do big blowups
> that will have people gazing intently at them.
> 
> -- 
> (This computer runs FC7, my others run FC4, FC5 & FC6, in case that's
>  important to the thread.)
> 
> Don't send private replies to my address, the mailbox is ignored.
> I read messages from the public lists.
> 
Jpeg is a horrible compression algorithm.  I have no idea why it was
chosen.  It is lossy, has artifacts in the result, often smearing the
image or showing a fingerprint like Moire pattern (most image processing
software has an anti-moire filter specifically tailored for JPEG ) and
other artifacts.  There are better compression algorithms, but many if
not all are proprietary.  Somewhere in Europe about 2005, there was a
new lossless algorithm similar to LZ in compression size or slightly
better, and HP licensed it, but I don't know the name.  LZ and GIF are
also lossless, but GIF somehow lost the ability to do color from what I
understand.  The problems endemic with images deal with the
representation, and the problem with something called tilt, which is an
artifact of the digital image capture method from CCD's.  CMOS devices
don't exhibit this or at least don't have to because the underlying
technology is different.

	CMOS is becoming the imaging technology of choice, but CCD's will be
around for a while.  The CCD's also suffer more from blooming, (for
those of you with film experience, think of acetate contamination) where
intense light will affect the surrounding pixels.  CMOS doesn't
experience this from the base technology although a poor screen which is
the surface that covers the detectors may cause a similar effect from
light diffusion.

	The number of megapixels to replicate 35mm film is dependent on several
issues, from the size of the imager, to the depth of field of the
lenses, and the formation of the screens, along with the design of the
technology.  However, for a normal 1/4 to 1/3 inch die, 15MP seems about
right.  For larger imagers, such as the focal plane imagers for 35mm
cameras, somewhat more mp may be required.  Cannon's current top model
now offers the EOS-1Ds Mark III with 21Mp.  For still more resolution,
the high end manufacturers like Hasselblad offer large format digital
SLR's with larger image sensors and >30Mp right now. Hasselblad's H3D II
for example offers a sensor of 48x36 and 39Mp.  which yields a 44
nanometer pixel size for capture, and the lenses are standard Hasselblad
system lenses, so if you put on a telephoto with a 5 degree field of
view (about 1000mm lens), you would get a pixel covering 641
nanodegrees, with a dynamic range of about 108db (film is typically only
about 65db).  
	
	The problem comes in storing and manipulating raw images of this size.
However the recent introduction of 8-32Gbyte flash is going to make this
technology more palatable.  Add the ability to fully seal the camera,
which will reduce dust and other problems with long term use, and the
writing is on the wall about film.  
	
	In addition, using encryption, and internet, a professional's images
can be anywhere in the world in seconds.  Even theaters are starting to
join the digital revolution, with fully solid-state projectors.  For
some examples look up Texas Instruments DMD projectors and chips.  There
is also a single pixel camera using a dmd device to scan the focused
image.  For still or slow speed, this would permit really flat images
with no pixel difference issues at all, very good for some kinds of
scientific studies.  Additionally there are highly optimized scientific
sensors capable of looking from the very low infrared to the very high
ultraviolet regions and beyond that are finding niches in microscopy,
biology, satellite, astronomy and other fields of research.  Some of
these images cannot be captured with any existing film due to
sensitivity issues, changes in required focus (look at your camera if
you have an slr.  See the red mark on the focus barrel?  That is for
infrared film.) and other issues (like getting the film back from
space).

	Some other sensors offer amplification of the detected image, which is
similar to low light image intensifiers with perhaps less capability,
but fewer problems and greater reliability.

	Most of the complaints about digital images are actually a result of
the post processing that most consumer cameras do on the image.
However, if you have a camera that offers it, try capturing the raw
image and bring it into a high quality bit of photo processing software,
and check out that result.  I think you will be surprised.

	A full sized 8.5x11 image at 3Mp would give you a pixel size of
7.72microns (7.72e-6 inches) in area.  This translates to about 360
linear dpi.  At 3' the average human eye can only see 3/1000" errors.
at 12" that would be 1/1000" errors, and so this image would be on the
edge of what you could see.  To get this quality, you must use photo
quality jets, ink, paper and your printer must be in tip top condition.
I don't know the latest on Laser stuff, but you can look it up.

Regards,
Les H




More information about the fedora-list mailing list