Color demosaicking using direction similarity in color difference spaces

In this paper, we propose a color demosaicking algorithm which adaptively selects direction for interpolation based on similarity of the directions in the color difference spaces. As color differences are usually smooth along the edge, we propose to estimate missing channels by selecting a direction which has higher score in similarity measurement. The complexity of our proposed method is low that only addition, subtraction and shifting are required. Experimental results show that our proposed algorithm has relative high PSNR and improvement in the visual quality of reconstructed images comparing to that of state-of-art demosaicking algorithms.