Bayer Colour Filter
white paper
In general single-sensor colour cameras are employing a monochrome
sensor with a colour filter pattern (one other possibility
to achieve a colour image with only one sensor would be to
use a revolving filter wheel in front of a monochrome sensor,
but this clearly has some limitations).
This method of achieving colour images has as a consequence
that no object point is projected on more than one sensor
pixel, that is, only one measurement (for a single colour
or sum of a set of colours) can be made for each object point.
In the following some different popular filter arrangements
are detailed:
Bayer Colour Filter (Primary
Colour Mosaic Filter)
The following table shows the arrangement of the filter pattern
for a sensor of size xs x ys (xs and ys being multiples of
2):
| |
0 |
1 |
2 |
3 |
.. |
xs-2 |
xs-1 |
| 0 |
|
|
|
|
|
|
|
| 1 |
|
|
|
|
|
|
|
| 2 |
|
|
|
|
|
|
|
| 3 |
|
|
|
|
|
|
|
.. |
|
|
|
|
|
|
|
| ys-2 |
|
|
|
|
|
|
|
| ys-1 |
|
|
|
|
|
|
|
As can be seen the building block is a 2x2 pattern with 1
blue and 1 red but 2 green filters. The reason for this is
that the human eye is more receptive to green than to the
other colours. Also this arrangement is less sensitive to
aliasing effects of more regular filter arrangements such
as the primary colour vertical stripe filter (explained below).
Most digital consumer cameras and camcorders use the Bayer
colour filter arrangement or the complementary colour
variant (explained below).
Complementary Colour Mosaic Filter
The following table shows the arrangement of the filter pattern
for a sensor of size xs x ys (xs and ys being multiples of
2):
| |
0 |
1 |
2 |
3 |
.. |
xs-2 |
xs-1 |
| 0 |
|
|
|
|
|
|
|
| 1 |
|
|
|
|
|
|
|
| 2 |
|
|
|
|
|
|
|
| 3 |
|
|
|
|
|
|
|
| .. |
|
|
|
|
|
|
|
| ys-2 |
|
|
|
|
|
|
|
| ys-1 |
|
|
|
|
|
|
|
This is basically the same arrangement as before except
without primary colours (R, G, B) but rather with complementary
colours (magenta, cyan, yellow). The reason for this is
that a primary colour filter blocks off 2/3 of the spectrum
(i.e. green and blue for a red filter) while a complementary
filter only blocks off 1/3 of the spectrum (i.e. blue
for a yellow filter). So the sensor is 2 times more sensitive.
The drawback is a somewhat more complicated computation
of the R, G, B values requiring the input of each complementary
colour.
Primary Colour Vertical Stripe Filter
The following table shows the arrangement of the filter
pattern for a sensor of size xs x ys (xs being a multiple
of 4):
| |
0 |
1 |
2 |
3 |
4 |
5 |
.. |
xs-3 |
xs-2 |
xs-1 |
0 |
|
|
|
|
|
|
|
|
|
|
1 |
|
|
|
|
|
|
|
|
|
|
2 |
|
|
|
|
|
|
|
|
|
|
3 |
|
|
|
|
|
|
|
|
|
|
.. |
|
|
|
|
|
|
|
|
|
|
ys-2 |
|
|
|
|
|
|
|
|
|
|
ys-1 |
|
|
|
|
|
|
|
|
|
|
This arrangement is very simple and basically
suited well for machine vision applications. The drawback
is that the horizontal resolution is only 1/3 of the vertical
resolution.
|
