Demosaicking using geometric duality and dilated directional differentiation

This paper presents a new demosaicking algorithm which uses two cost terms: the interpolation error of a low resolution image based on geometric duality and the dilated directional differentiation of color differences. Since a given high resolution image and its low resolution image obtained by sampling have similar edge properties, the proposed algorithm computes the interpolation errors for the candidate directions in the low resolution image, and exploits them as a cost term for the direction. In addition, the interpolation direction can be determined accurately even in the vicinity of object boundaries by dilating the directional differentiation of the color difference values. Through dilation, some pixels, which are in the neighborhood of an edge but classified into a flat region by simple edge detection like the Sobel filter, are reclassified. By combining this edge classifier and the weighted sum of the estimates obtained by Taylor approximation, missing pixels are interpolated. Simulation results show that the proposed demosaicking algorithm is superior to other state-of-the-art algorithms in terms of visual and objective qualities. Furthermore, the computational complexity is comparable with the existing algorithms. Therefore, the proposed algorithm is suitable for real-time implementation.