Demosaicing for different Colour Filter Arrays

Description

Digital cameras are usually based on a single area sensor (CCD/CMOS sensor), with each pixel measuring the
intensity of the incident light at the corresponding point. In order to record more than just black and white,
colour filters are placed in front of the image sensor. This means that each pixel is assigned one of the
three RGB colour channels to each pixel. However, the colour information of the image is therefore incomplete because
two colour values per pixel are missing. Demosaicing is used to reconstruct the missing colour values.

The green colour channel is usually sampled with a higher resolution, as the human eye reacts most sensitive to differences in brightness in the wavelength range of the green colour.
One such typical colour filter is the Bayer filter, in which 50% of the pixels capture the green colour channel.

Left: Bayer filter, right: example of a non-regular quarter-sampling colour filter

 

 

 

 

 

 

Alternative colour filters, which need to be evaluated, find their approach in the image acquisition with non-regular sampling.
By non-regular sampling with subsequent reconstruction, aliasing effects can be minimised thus higher frequencies can be captured.
Such non-regular colour filter arrays are to be compared with the classic Bayer filter and different demosaicing approaches are to be evaluated.

Voraussetzungen

Experience in programming with MATLAB/Python and good knowledge of image and video processing is required.

Supervisor

For more details contact:

Teresa Stürzenhofäcker
teresa.stuerzenhofaecker@fau.de