Demosaicking for Fourier spectral filter array based on L1-norm minimization of high frequency

Abstract

We propose a novel demosaicking method for Fourier specral filter array (FSFA). The proposed method consists of 2 steps: spatial and spectral reconstruction. In the spatial reconstruction, we separately estimate low- and high-frequency components decomposed by a discrete wavelet transform. The low-frequency components are estimated by the bilinear interpolation, and the high-frequency ones are done by compressive sensing because of their sparsity. Experiment results shows that the proposed methods outperforms other existing demosaickings especially in the spatial reconstruction.