Dean wrote:
>The small file size will only occur for a subset of all possible images.
>Hopefully this subset includes the majority of photographic images. The
>best possible compression for an "image" that consists of nothing but
>random data is a one bit flag to indicate that the rest of the file is
>untouched. Luckily, most images are more interesting than random noise and
>compress with the appropriate compressor.
That's not bad, Dean, but a little hard to understand for the
non-programmer. May I take a shot at it?
Some photos are naturally more "random" than others. A medium head shot with
a background of tree-foliage will produce a *much* larger compressed file
than a full nude on a paper-roll background, for example. Architecture will
almost always compress better than a cornfield. Where compression is most
effective is where the colors in large portions are relatively flat, and
texture is at a minimum.
"Compression" starts breaking down when: 1)there is too much texture, or
2)colors are "flat" but not quite the same. In the first case, there isn't
very much compression. In the second case, artifacts creep in very quickly.
I'm sure some will disagree with me here, but I've seen it enough times to
say it with some confidence.
Best regards--LRA
>From: "Shough, Dean" <dean.shough@lmco.com>
>Reply-To: filmscanners@halftone.co.uk
>To: "'filmscanners@halftone.co.uk'" <filmscanners@halftone.co.uk>
>Subject: RE: filmscanners: (anti)compression?
>Date: Thu, 09 Aug 2001 04:57:53 -0700
>
> > >It turns out that it is impossible to create lossless compression
>scheme
> > >that does not cause some files to expand in size. A set of random
>files
> > >always expands. There is no way to encode the random information that
> > does
> > >not take up at least as much space as the original file. Because of
> > this,
> > >any image that contains lots of random noise tends to compress much
>less
> > >than a high quality image with little noise.
> >
> >
> > What about Genuine Fractals compression which claims non lossy
> > compression and small file size.
> >
>
>The small file size will only occur for a subset of all possible images.
>Hopefully this subset includes the majority of photographic images. The
>best possible compression for an "image" that consists of nothing but
>random
>data is a one bit flag to indicate that the rest of the file is untouched.
>Luckily, most images are more interesting than random noise and compress
>with the appropriate compressor.
_________________________________________________________________
Get your FREE download of MSN Explorer at
