At 1:34 AM +0000 9/21/03, olympus-digest wrote:
>Date: Sat, 20 Sep 2003 11:30:58 +0100
>From: "Julian Davies" <julian_davies@xxxxxxxxxxxxxx>
>Subject: Re: [OM] How many bits/pixel does a digital image need?
>
>I think this is true for capture and storage, but the lessons of audio tell
>us that synthesizing a couple of extra bits of noise below the data on
>reproduction can work wonders on areas of slow change. As long as there is
>sufficient data to represent the tonal range of the target film, the noise
>(dither) would represent the random grain of the film and smooth out blocky
>colour.
>I'm certain that our DSP pros can add much more to this than I can...
This intent is to provide enough bits so that quantization (and other kinds of
numerical) noise is insignificant compared to the physically inherent sources
of noise, so the system is limited only by physics. This requires a few more
bits than one would expect from the inherent noise alone, and is why many
high-end professional digital scanners capture 48 bits per 3-color pixel (16
bits per color), even whan the final output will have far fewer bits per pixel
per color. They do all the computations at full width, and truncate down to
the final pixel size at the last step only.
>- ----- Original Message -----
>From: "C.H.Ling" <chling@xxxxxxxxxxxxx>
>To: <olympus@xxxxxxxxxxxxxxx>
>Sent: Saturday, September 20, 2003 8:22 AM
>Subject: Re: [OM] How many bits/pixel does a digital image need?
>
>
> > Having been immersed in digital imaging (from scanner to digital camera) for
> > ten years I don't see deficiency in Z-axis of today's prosumer digital
> > camera (E-10 and 10D), they are indeed better than slides not to mention
> > prints which only has a D-range of around 2.0.
A D-range of 2.0 (100:1 linear range) and 12-bit pixels (per color) yields a
signal-to-noise ratio of (2^12)/(100)= 40.96 (assuming that the camera is
quantization-limited, which seems to be the case for prosumer cameras, despite
the above analysis)). Now, 40:1 is very good, as you observe, but is not quite
in open-window territory.
Someday I'll get a digital camera, and will inflict upon it the most intrusive
of tests, to learn what exactly limits this unfortunate camera. I'll bet that
this kind of information is closely held by the camera manufacturers, and even
if the photo magazines know of these tests, they probably cannot see how to
explain the meaning of such tests to the mass-market consumer.
Joe Gwinn
< This message was delivered via the Olympus Mailing List >
< For questions, mailto:owner-olympus@xxxxxxxxxxxxxxx >
< Web Page: http://Zuiko.sls.bc.ca/swright/olympuslist.html >
|