Grayscale perception
Grayscale perception
- Subject: Grayscale perception
- From: "Mark Rice" <email@hidden>
- Date: Mon, 5 Dec 2005 17:52:09 -0500
- Organization: Zero One
Marco -
It is true that there are an infinite number of points along a line, but in
16 bit, there are only 65,536 points on the line. Even that is deceptive, as
almost all 16 bit files are converted to 8 bit before being printed. Usually
the point of having a 16 bit file is to have enough data points to convert a
curve of some sort, whether it be logarithmic or some other shape, to a
linear file without quantization.
So what we really need to ask is, can the eye detect more that 256 shades of
gray? The answer is yes - usually an 8 bit (256 shades of gray) will show
some banding, unless dithering is introduced.
Different people can see considerably different amounts of discrimination,
so it is impossible to answer the question: "What different values can
people see?"
One measure of this is the Munsell color test. I took it when I got my first
job in the photo lab business to see what my color discrimination was. It
consisted of 256 color chips, mounted on wooden circles. The examiner mixed
them up, and then asked the examinee to put them in order. They used subtly
different hues - light desaturated green to light desaturated cyan, for
instance. The differences between colors were VERY subtle.
I think the only way to truly answer the question would be to give a large
group of users a gray scale discrimination test similar to the Munsell test,
and tabulating the results. It would also be necessary to generate a series
of smoothly gradated gray values, as the eye may be more sensitive to smooth
changes than it is to discreet samples. I do not know how one could create
the latter test, as I know of no printers that can generate prints with more
than 8 bits without dithering.
Mark Rice
email@hidden
www.zero1inc.com
In a message dated Sun, Dec 4, 2005 11:54 PM, Nathan Duran wrote:
>> Can humans really perceive such slight variations in density, and if
not,
>> how can there be lab values for such distinctions?
>
> There are an infinite number of points along any line. Whether or not any
> human can distinguish them all by sight doesn't really matter to the
> numbers.
I agree. We are dealing with mathematics as applied to human perception.
>From my limited understanding, it seems obvious that mathematical precision
requires that many steps in 16 bits.
Whether or not these steps are perceptible is relatively unimportant,
whereas what matters is that the calculation is effected with the maximum
possible precision. From there, one can always proceed to rounding numbers
off, if needed.
Can anyone please either confirm my impression, or rebut?
Thank you.
--------------
Marco Ugolini
Mill Valley, CA
_______________________________________________
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