Главная страница | назад





Article #15972: Color reduction techniques for graphics.

 Question and Answer Database
FAQ972D.txt Color reduction techniques for graphics.
Category :Windows API
Platform :All
Product :All 32 bit
Question:
How do I accurately (and automatically) display HiColor and
True Color (16 and 24 bit) images on a 256 color display?
Answer:
Generally, you cannot. This requires sophisticated color
reduction techniques, and not all will work well on all images.
Most all color reduction methods work by weighing the average
of similar colored pixels in a given are of the image, and
combining them. Consider a image of a person standing in a park
with green grass, flowers, trees and a blue sky. You would need
to remap potentially millions of colors to fit into the Windows
palette (236 colors). While this can usually be done in a
professional paint program with some trial and error, a single
function would have much trouble remapping the colors to
faithfully reproduce all images. This is the reason why paint
programs that can accurately reduce colors give so many choices
in the method used to reduce the color.
In the above example, you could combine many shades of blue and
green, and dedicate more colors to faithfully reproduce the
persons flesh tones. This method would probably produce an
acceptable image, since the shades of the sky, grass, and trees
are far less important than the shades necessary to reproduce
flesh tones. On the other hand, if the purpose of the image was
to show a spectacular panorama of a canyon, you might want to put
less emphasis on the exact color of a house that might be in the
picture, and instead concentrate on the colors of the canyon. As
you can see, where one color reduction function may work well on
one image, it would most likely have trouble with others.
Some paint programs use a "Color Wash" palette, that sometimes can
display with fair accuracy a colorful 24 bit photograph. This is
done by creating a palette that contains a wash of colors that are
generally useful for displaying most images. Again, pixels are
compared by area, and remapped to a color that is reasonably close
to the original if possible.
7/16/98 4:31:28 PM

Last Modified: 01-SEP-99