Re: Decimal precision in Photoshop's info palette
Re: Decimal precision in Photoshop's info palette
- Subject: Re: Decimal precision in Photoshop's info palette
- From: Karsten Krüger <email@hidden>
- Date: Sun, 21 May 2006 14:06:28 +0200
Am 21.05.2006 um 12:26 schrieb Dennis W. Manasco:
Let's argue for floating-point representation of color at every
stage in PhotoShop (including dialog entry), except for output to
other formats and printing.
That's a movement I can get behind.
Unfortunately, I'll probably have better luck tilting at windmills...
Be happy for the windmills...
If your wishes come true PS has to encode the whole bitmap in 128 bit
real - which means your images will be very huge:
8bit means 8bit per color - 24bit in RGB, 32 bit in CMYK
16bit : 48 bit in RGB, 64bit in CMYK
128bit : 384bit in RGB, 512 bit in CMYK
This is essentially 16 x the space occupied in RAM and on HD. 10 MB
8bit picture will become 160MB, without adding any information.
For number crunching it will be more than 16 x the amount of time it
takes to do calculations today. All math today is done using highly
optimized integer code. Each operation can be done inside the chip
using one or two registers which is highly efficient. The whole pixel
can be manipulated in one processor cycle. On modern microprocessors
this is true for 16bit colors, too, because registers are 64 bit
wide. With 128bit real all color panes have to be treated separately
for each pixel. Cross-processing color information can not be done
executing simple matrix operations because of this, too. A filter,
which takes about 10 seconds to execute today, will need at least 2.6
minutes for the same task in 128bit world. This is not practical on
today's computer hardware.
And why do you want to do so, anyway ? A standard offset machine is
able to print about 300.000 distiguisable colors, add Pantone and you
max out at 500.000 colors. A high end CRT has about 1.5 mio
distinguishable colors, highend LCDs coming close to 1 mio. So your
8bit graphics card is able to drive 16.7 mio colors, but only 10% of
them are distinguishable on your CRT. And about 1/3 of what you see
on your CRT is distinguishable in offset printing. So even 8bit is
more than enough for direct image representation.
Problems arise when moving between gamuts or stretching them (like
pushing an underexposed image to gain more details in shadows or
removing a color cast) or due to some non-linearities. But these
problems can not be solved by extended bit depth. These are
limitations of the color model being used (cmyk/rgb relative to some
limited primaries), hardware issues (like non-printable colors) and
how to compensate them.
Anyway, have fun with your scripts,
Karsten
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Colorsync-users mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden