Digital Colour Imaging Techniques for Machine Vision Cameras

All machine vision camera sensors are inherently monochrome and colour information is obtained either by a colour mosaic filter placed over the sensor or the light is separated by a prism into its red, green and blue colours before being directed to three different sensors.

There are two types of colour mosaic filter techniques – the Bayer Colour Mosaic Filter and the Complementary Colour Mosaic Filter.

Bayer Colour Mosaic Filter

The Bayer Colour Mosaic Filter involves a sensor with a sequence of red, green and blue filters placed over the top. In this mosaic filter there are generally twice as many green pixels as blue or red to mimic the human eye. To achieve a full colour image, interpolation is performed to provide a red, green and blue value for each pixel. This can be done within the camera or externally.

Complementary Colour Mosaic Filter

The second type of mosaic filter is a Complementary Colour Mosaic Filter which is used where low light sensitivity is required. The filter is composed of yellow, cyan, magenta and green pixels and the filter is less dense allowing more light to pass through.

The way the pixel colours are interpolated varies depending on the colour of the pixel. If the red or green values for a blue pixel are required, the four of the surrounding pixel values are used. If the red and blue values are required for a green pixel, two surrounding pixel values can be used.

The three chip colour imaging technique uses a prism behind the camera lens to split the light into its component colours and a separate CCD sensor is used for each of the red, green and blue spectrums. The images are output from the camera separately and recombined during processing. In some cases the images can also be combined in the camera and output as S-Video, composite colour or YUV.

The colour mosaic filter technique is generally lower cost than the three chip solution. It is also easier to choose a suitable lens as the three chip method requires a closely matched lens which does not protrude into the body of the camera. Standard lenses will often result in chromatic aberration where abnormalities in the lens refract light differently. This in turn leads to spatial separation between the planes.

The bandwidth in a mosaic filter camera can be reduced by converting the Bayer to RGB values on the host PC. Typically mosaic filtering is more sensitive than a three chip camera, however the conversion to RGB can be processor intensive in real time and the filtering also results in less colour fidelity. Mosaic filtering can also introduce distortion when viewing lines and edges in some cases, due to lack of information.