Image demosaicing using content and colour-correlation analysis

• We propose a content adaptive demosaicing strategy utilising structure analysis and correlation between the red, green and blue planes.
• The filter coefficients are optimised through an offline training procedure.
• The online filtering process is very efficient and performs comparatively with much more expensive methods.

Most digital cameras of today use a single CCD image sensor with alternating arrangement of red, blue and green colour filters in what is known as a Bayer pattern. To extract a full colour image, a demosaicing strategy has to be applied. In this paper we propose a content adaptive demosaicing strategy utilising structure analysis and correlation between the red, green and blue planes. These two aspects are used for the classification of a block of pixels to generated trained filters. The proposed method aims to reconstruct a high quality demosaiced image from a Bayer pattern in a colour filter array efficiently. Experimental results show that the proposed strategy performs comparatively as more expensive methods.