1993-12-01 - Re: Statistics of Low-Order Bits in Images

Header Data

From: Timothy Newsham <newsham@wiliki.eng.hawaii.edu>
To: jazz@hal.com (Jason Zions)
Message Hash: a958c67d93cf153cf88a362f80aef44de3e46bda3913350b311705ec6ead7724
Message ID: <9312010332.AA17576@toad.com>
Reply To: <9312010256.AA02876@jazz.hal.com>
UTC Datetime: 1993-12-01 03:32:51 UTC
Raw Date: Tue, 30 Nov 93 19:32:51 PST

Raw message

From: Timothy Newsham <newsham@wiliki.eng.hawaii.edu>
Date: Tue, 30 Nov 93 19:32:51 PST
To: jazz@hal.com (Jason Zions)
Subject: Re: Statistics of Low-Order Bits in Images
In-Reply-To: <9312010256.AA02876@jazz.hal.com>
Message-ID: <9312010332.AA17576@toad.com>
MIME-Version: 1.0
Content-Type: text/plain

>    >Sounds like the simple solution is for people everywhere to replace
>    >the low order bits of all of their pictures with good random noise. 
> You're kidding, right? Image processing software, and music processing
> equipment, are designed to get the maximum performance out of the storage
> medium. I find it pretty unlikely that any developer of image processing
> tools is going to deliberately drop one bit of precision off each n-bit
> unit; sheesh, image files are big enough without software deliberately
> wasting bits.

data is most usually stored in a way that makes sense on the underlying
hardware, ie 24 bit palette.  Why 24? because it is 3 bytes per pixel,
one byte for R,G and B.  R G and B arent even weighted the same perceptually!
Why should they have the same amount of bits? Because it makes sense on
the machine!

> As for introducing noise into the low-order bit of music CDs, definitely
> not; the trend is towards equipment that can extract the last dribble of
> information from well-recorded sources, and towards recording equipment that
> can meaningfully record every bit of every word. Why do you think there are
> 64x-oversampling players and the like?

Not everyone is trying to squeeze every last "bit" of performance,
take for example crystal clear fone lines over fiber optic connections
with extra noise added in to sooth the listener.

I'm not familiar with oversampling in CD players but where I am familiar
with anti-aliasing is in cases where it is easier to do anti-aliasing
filtering digitally than it is to do it with analog circuitry.  Maybe
you can tell me why there are 64 times oversampling playres.

> Eric's right; you'll have to build your own tools to make things noisy, as
> the trend elsewhere is to make things unnoisy.
> On the other hand, have you looked at the new mini-disc technology? That
> stuff already introduces audio compression, and the bits on a minidisc
> should be pretty high in entropy; of course, twiddling bits in a compressed
> audio recording may have dramatic effects on sound...

The compression in these mini-discs is more of a coding scheme than
a compression I believe.  It eliminates elements in the sound that
are perceptually unimportant to the listener.

> Jason