Charles Robinson wrote:
On Dec 27, 2005, at 12:32, Dario Bonazza wrote:
Good point, hence I'm no longer sure of my previous idea. On the
contrary, I'm afraid you are right and I need someone else's help
now: can we take for granted that any pixel is always present within
the histogram?
Probably... but you have to remember that the histogram is a
distribution graph. If there are only one or two "full white" pixels,
they are not going to make a big tall line on the right-hand side. You
need a fair number of pixels (I'm dodging the question here, as I don't
know really HOW many are required) to get a noticeable line to show at
a specific point on the graph.
I think the graph is probably normalized, i.e relative and not absolute.
The reason I think this is that some histograms with a very even
distribution have a very large (in area) curve, whereas extreme
distributions many pixels with a very narrow range of luminance, have a
much smaller area under their curve.
Therefore there is no "minimum" that will trigger a dot on the display,
it all depends on the distribution.
-Charles
--
Charles Robinson
[EMAIL PROTECTED]
Minneapolis, MN
http://charles.robinsontwins.org
