on 17/11/04 11:59 am, matthew ward wrote:
>
> ...
> Even Photoshop lies about bit depths.
> It uses a rather odd 32,769 levels!
> Maybe (for legal purposes) they are claiming that 15-bit +1 is the
> same as 16-bit?

The file formats use 16 bits of storage space and computers efficiently
manipulate 16 bits (or 8 or 32). There are no 15-bit standard file formats,
and computers are much slower at extracting and manipulating 15 bit values
than 16 bits. The reason that Photoshop uses pixel values of 0-32768 (which
requires 16 bits but doesn't use the possible values from 32769-65535)
instead of 0-32767 (which would fit in 15 bits) or 0-65535 (which fits in 16
bits) is this: 0-32768 has an exact middle value; 0-65535 does not. A lot of
the blending and other math that Photoshop does can be done more efficiently
if there's a middle value.

In 8-bit mode, pixel values are 0-255. Blend modes like overlay and soft
light that have a neutral color of 50% gray have to actually use a value
either slightly lighter or darker than neutral gray because there is no
value exactly in the middle. Then that has to be taken account of in the
math.

When 16 bit support was added, the loss of ability to represent values from
32769-65535 was felt to be a worthwhile tradeoff for the increase in speed
made possible by having a representation with an exact middle value. 16 bit
is already slow enough, and there's almost no source of data (camera,
scanner, etc.) or sequence of useful image manipulations that need those
extra values. Anything that needs more than 0-32768 probably needs some kind
of high dynamic range.

Bottom line: it's an engineering tradeoff, not a legal issue, and not
Adobe's attempt to cheat you out of a bit that you've paid for.

Russell


===============================================================
GO TO http://www.prodig.org for ~ GUIDELINES ~ un/SUBSCRIBING ~ ITEMS for SALE

Reply via email to