Color filter array interpolation by successive refinement over color channels using gradient inverse-weighted filtering

Abstract In this paper, we present a novel color image demosaicking algorithm based on a directional weighted interpolation method and gradient inverse-weighted filter-based refinement method. By applying a directional weighted interpolation method, the missing center pixel is interpolated, and then using the nearest neighboring pixels of the pre-interpolated pixel within the same color channel, the accuracy of interpolation is refined using a five-point gradient inverse weighted filtering method we proposed. The refined interpolated pixel values can be used to estimate the other missing pixel values successively according to the correlation inter-channels. Experimental analysis of images revealed that our proposed algorithm provided superior performance in terms of both objective and subjective image quality compared to conventional state-of-the-art demosaicking algorithms. Our implementation has very low complexity and is therefore well suited for real-time applications.