Date   : Sun, 08 Jul 2001 11:18:25 +0100
From   : Russell Marks <russell.marks@...>
Subject: Re: Magazine test scans - Update

(Apologies for dragging this out on the mailing list, but I couldn't
bring myself to let that post stand unchallenged.)

Paul Wheatley <p.r.wheatley@...> wrote:

> Russell Marks wrote:
> > 
> > Paul Wheatley <p.r.wheatley@...> wrote:
> > 
> > > This is only an issue if you don't set the quality of compression high
> > > enough. If you want to use JPEG as a handy, widely used, open format then
> > > simply compress all images on 100 which makes them lossless. They will
still
> > 
> > Not true. From the JPEG FAQ:
> > 
> > > Except for experimental purposes, never go above about Q 95; using Q 100
> > > will produce a file two or three times as large as Q 95, but of hardly any
> > > better quality.  Q 100 is a mathematical limit rather than a useful
setting.
> > > If you see a file made with Q 100, it's a pretty sure sign that the maker
> > > didn't know what he/she was doing.
> 
> This isn't true either. I've worked pretty seriously with graphics for
> several years now (having coded things like Delirium on the RiscPC). I've

Ok, so you wrote a screensaver, and I've written a picture viewer. Are
we even now? :-)

> component of JPEG compression. However, in my experience a Q100 is virtually
> identical to the uncompressed original if not exactly the same. Certainly

You're missing the point. When talking of compression algorithms,
"lossless" has a very specific meaning - that you can get bit-for-bit
identical output from the decompression. (Which you can find
explicitly mentioned in the JPEG FAQ, BTW.) So your claim that using
quality 100 for a JPEG makes it lossless is false.

> anyway). Q100 *is* useful if you want to use the JPEG format (because of its
> common use and wide support) but don't want loss of image.

The usefulness or otherwise of Q100 notwithstanding, the "loss of
image" is demonstrably present:

rus@.../home/rus/tmp>pgmramp -ellipse 640 480 >a
rus@.../home/rus/tmp>cjpeg -q 100 a |djpeg >b
rus@.../home/rus/tmp>cmp a b
a b differ: char 265, line 4

This is on the same platform, with the same JPEG implementation, with
an integer DCT, with a mindlessly simple (yet JPEG-friendly) artifical
image. Surely optimal conditions. And it's *still* not lossless.

>                                                            The guy that wrote
> the JPEG FAQ really doesn't know what he's talking about.

The author of the JPEG FAQ, Tom Lane, is also one of the main authors
of the Independent JPEG Group's JPEG software, as used by most
programs which read/write JPEG files.

So, unless you have some pretty compelling evidence, I think you might
want to retract that.

-Rus.