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: 8bits vs. 16bits/channel: cantheeyeseethedifference

Hi Roy,

> > This sounds like in going from 48 bit color to 24 bit color, you're
> > losing spatial resolution. I never realized that before. That's what PS
> > actually does? Of course that IS dithering, not aliasing. Dithering is
> > something engineering deliberately does. Aliasing is something
> > engineering tries to deliberately avoid.
> >
> Yes, PS really does do this.

Do what?  Decrease spatial resolution when converting from 16 bit space to 8
bit space?  No.  Absolutely not.  It decreases TONAL resolution, not
spatial.  Entirely different thing.

> It isn't too hard to demonstrate.
> (if you are curious enough I'd be glad show this).

Please.  I'm going to assume you really don't agree that it is decreasing
spatial resolution, but decreasing tonal resolution...

> But don't worry about the resolution at all.  I'm only talking about
> differences less than the least significant bit

Can you give a concrete example of what you believe it does?  Do you believe
it takes 0x1289 and converts it to 0x13?  Let's say the entire image is
completely 0x1280...are you believing that ~half the 8 bit data is 0x13 and
~half 0x12?  Yes, if you believe that, and can demonstrate it, I'd like to
see it, and I would agree that is dithering.  But really, no entire field in
real life would consist of 0x1280...there would be a range, and in reality,
simply converting by lopping off the 8 LSBs will give you the same
"fidelity", from an analysis standpoint.

> The bad thing would have been to just truncate --
> in that case the average value would actually drop by 1/2 LSB.

But no matter what you do, it's always going to drop 1/2 LSB...

> If you have a value of 127.5 is HAS to go to either 127 or 128 so
> it makes sense to split the difference.   If the value was 127.3
> 30% would go to 128 and 70% would go to 127 -- much better than
> just truncating and even better than always rounding as well.

I disagree.  If you were dealing with pure data, yes, but this is image data
that just doesn't have entire fields of 127.5 values...so even if it did as
you suggest, it would, at best, increase the fidelity insignificantly in a
real world image.

First, show me that it in fact does as you believe.  Then we can discuss the
significance.  It's real simple to simply take a 16 bit image and just lop
off the lower 8 bits, and then "compare" that to an image that has been
"processed" as you suggest, and see if there is a noticeable difference.



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.