Content dependent demosaicing algorithm implementation

In this paper, the novel demosaicing algorithm based on high-level features of human visual system (HVS) is proposed. Fixation density maps (FDM) obtained by regions of interest (ROI) analysis are used to separate areas in the picture that are important for the observer (foreground) from the unimportant ones (background). Two known interpolation methods with different performance were employed for reconstruction of foreground and background. Performance of the proposed algorithm is evaluated using objective and subjective quality tests and compared to the state-of-the-art techniques.