Apache-Talk @lexa.ru 

Inet-Admins @info.east.ru 

Filmscanners @halftone.co.uk 

Security-alerts @yandex-team.ru 

nginx-ru @sysoev.ru 




      :: Filmscanners
Filmscanners mailing list archive (filmscanners@halftone.co.uk)

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[filmscanners] Re: JPEG2000 > Paul

Paul wrote:
>I wouldn't obsess too much about lossless compression. If I take an image
>file of some bright blue cloudless sky, either from a digicam or a scanned
>slide, and look at the numbers, they have up to five units (peak-to-peak) of
>noise in them. That is, the blue value may jump around between, say, 180 and
>185 in a very small area that looks solid to the eye. After saving an
>reloading an image in a lossy format, some pixel values may be off by one or
>two. That's therefore an utterly insignificant loss. By tolerating that
>small additional error, you can get 10:1 compression instead of 2:1.

Except that

a) it matters WHERE the modified pixels are, and how they are distributed.
If distributed in patterns these can be detectable much more easily than if
diffuse.  This is where jpeg has a problem, with its visible jiggles around
transitions. I understand from what you and Robert say that j2k puts the
changes that occur with compression in a more widely spread manner, so they
are not so visible.  My own experiment on 10% files did not show much
difference in overall quality betw jpeg and j2k - they were different in
appearance, but not necessarily one better than the other. I need to try
with other images when more time.

b) contrary to your experience, when I looked at the histogram of the
difference between original and compressed images, I saw  a bell curve of
"difference pixels" differing from the original by around 4 levels, but
also a 'tail' of lesser numbers of pixels differing by up to 30 levels
(j2k) or 37 (jpeg) from the original.  There were not so many of them, but
they would be much more visible.  Again it depends on how they are
distributed as to how visible they are.

I am only nitpicking (or is it investigating?) because it is a big decision
to decide to process all your images (including most especially the larger
16-bit raw scans) in a way that will definitely compromise them - the
question being how much?

Maybe there is a fault in the Fnordware j2k plugin that I downloaded which
means the results are not as clean as they should be?


Canberra, Australia

Satellite maps of fire situation Canberra and Snowy Montains

Unsubscribe by mail to listserver@halftone.co.uk, with 'unsubscribe 
or 'unsubscribe filmscanners_digest' (as appropriate) in the message title or 


Copyright © Lexa Software, 1996-2009.