A true colour digital image requires 24 bits to specify the
colour of each pixel on the screen. It is expensive to
have a high-speed memory to support such a full-colour
display on a high resolution display. An alternative
solution is to provide a limited number of bits for
specifying the colour of each pixel. Each of these values
is then used as an index into a user-defined colour palette
table. The process of carrying out a colour palette table
selection is known as the colour quantization process.
The colour quantization is one of the most useful lossy
compression methods [1, 2] with the benefit of easy
implementation on the image receiver site. Typically, the
colour quantization attempts to find an acceptable set of
palette colours that can be used to represent the original
colours of a digital image [1]. There are two points that
make this lossy image compression successful. First of
all, it exploits the limited ability of human perception
which is capable of distinguishing less than a thousand
colours. In addition, it exploits the limited capability of
many display devices to display true colours on many
display devices like a billboard [3] or a low quality LCD
(Liquid Crystal Display) which have constantly been
used as replacements for text centric display media.
Applying the colour quantization to an RGB colour
image frame of size 640 9 480 pixels, the resulting
image with a colour palette size of 256 will have only
(640 9 480) ? (256 9 3) = 307,968 bytes which is
about one-third of the original image size. This makes
the colour quantization process widely utilized in many applications especially in computer graphics and image
processing.