Gamma 1.0
Gamma 1.0
- Subject: Gamma 1.0
- From: "Tom Lianza" <email@hidden>
- Date: Tue, 25 Sep 2001 08:15:53 -0400
>
Hey there-
I'd like to hear peoples' opinions on using an RGB space with a gamma of
1.0.
<
When I saw this question one person's name came to mind: TIMO
Now there's an individual that has demonstrated a real lack of understanding
about color IN WRITING.
Some of what he says is correct, but his outrageous assertions
tend to get in the way of what ever message he's trying to put forth.
He posted some outrageously incorrect information on the sci.engr.color news
group and routinely
gets into some nasty conflicts with Poyton. You wouldn't want to confuse
either of these guys with a color scientist, but Poyton generally has the
humility to point out that he is a good writer and a good listener and he
generally runs his work past collegues that may be more in tune in the
sciece
In general, there are certain operations that lend themselves to maintaining
linear data. There are very well founded arguments
which relate to the areas of quantization artifacts and to the very nature
of certain operations that work best in a linear space.
For instance, general sharping and bluring functions that rely upon a
mathematical principle of convolution, assume a linear intensity
relationship. A sharpening function such as unsharp masking relies upon a
logarithmic definition of the data. The most fundamental
problem in all of this is that one has maintain linear data at a very high
resolution throughout the process and there is very little to gain
from it in practice, PARTICULARLY if you wish to display the data on a
monitor or output onto a hardcopy device. Nearly all
hardcopy devices have native gammas that are greater than one so at some
point you have to apply gamma correction. I've done a number of
scanner designs (the first in 1977) and we traditionally quantize the data
to high resolution within the scanner and then apply the gamma correction to
map that to an 8-bit data space. If you start with atleast 10 bit data
(linear) and then map it to a gamma of approximately 2.2 there is generally
no loss of codes. If we start performing many different types of
operations on the gamma encoded data, TIMO is quite correct that there is
plenty of opportunity for loss of quality. You can do incredibly stupid
things in Photoshop to permently damage the final image. The use of
adjustment layers in Photoshop 6.0 overcomes the bulk of these problems that
occur when consecutive operations are applied to an image.
The use of a gamma 1. workflow requires that one maintain enourmous amounts
of data which is ultimately lost in the final output step. It requires a
great deal of rendering capability to simply display an image on a CRT/LCD
without introducing quatization on output. For 99.9% of the users it
introduces more potential for error than any potential gain in percievable
quality. On the other hand for certain arcane operations, it is useful.
Take care and good luck....
Tom Lianza
Technical Director
Sequel Imaging Inc.
25 Nashua Rd.
Londonderry, NH 03053
Tom Lianza
Technical Director
Sequel Imaging Inc.
25 Nashua Rd.
Londonderry, NH 03053