Patrick Elliott wrote:
> No they wouldn't. They would only see the ones produced on systems with 
> brighter or darker displays that way. Most, since most people are using 
> basically the same hardware, don't play games, are not pros when it 
> comes to graphics, etc, **never** adjust the brightness of the display, 
> so unless the guy sending the image is on a Mac and you are viewing it 
> on a PC, no difference is seen at all. Even if this does happen, most 
> people chalk it up to the image "looking as intended", even if it isn't, 
> they just don't know any different. Its only when two people with 
> different systems actually compare notes that it becomes obvious that 
> the PC made image or Mac made one is bleached or too dark. If you don't 
> know there is a problem, you won't notice the problem, and sadly, since 
> most applications do Software gamma, instead of hardware correcting by 
> increasing/decreasing the entire output to the display, you can actually 
> get "corrected" images that look too dark of bleached, because the 
> colors on them "can't" be corrected numerically. How do you make &FEFE60 
> "brighter" without making it almost white, or &101000 darker, without 
> making it almost black? You can't. You can "hardware correct" the entire 
> display to make "everything" brightened or darker, without altering the 
> actually values in the image.
> 
> Frankly, other than the existing issues, I think there is a more serious 
> problem with Gamma. Mainly, to work right you would need a graphics card 
> that supported "more" colors than actually possible, in other words, use 
> 8 bits per channel for the "image", but have 10/12 bits per channel for 
> the "card". That way the "image" would reside in the "middle range", but 
> color values on the card might range from -15 to 270 per channel 
> (forgive me if I did the math wrong). That way, if you need to gamma 
> correct you *shift* the values of the image, not the entire display. I 
> am 99.9% sure that it doesn't work that way, so you either end up 
> adjusting the "values" in the image, which won't work if they are too 
> close to the white or black thresholds, *or* you adjust the entire 
> display to compensate, which then messes up any windows you are "not" 
> viewing the image in.

I think you have slightly misunderstood gamma correction. The blackpoint 
and the whitepoint are not affected by gamma, it is only the middle 
range of values that are altered. 0 will be 0 and 255 will be 255 
whatever the gamma, no need for values "out of range". But there is a 
problem with the limited number of discrete values of 8 bit numbers if 
gamma correction is done in software.
To take your examples: FEFE60 gamma corrected to be brighter might end 
up being FEFE72, 101000 might become 090900 (I did the correction 
from/to 1.8 and 2.2). The values close to 00 and FF don't change much at 
all as you see, but there are some error introduced due to rounding. 
That error may show up as color banding.

You use the brightness and contrast knobs on you monitor to adjust the 
blackpoint and the whitepoint but that won't affect the gamma of the 
display. The color value to monitor output function is something like 
this (at least theoretically):
output = (brightness + contrast * value) ^ gamma
 From that formula it should be clear that there is no way to change 
brightness and contrast to compensate for wrong gamma and vice versa.
(A gamma knob on the monitor would be handy, maybe they have that on 
some monitors, I don't know)

-- 
Daniel Nilsson

Post a reply to this message