Spatial and Multispectral Correlations for Joint Design of Color Filter Arrays and Demosaicing
Gaoguo Qi, Zhenlong Du · 2020 IEEE 5th Information Technology and Mechatronics Engineering Conference (ITOEC) · 2020
CFA(Color filter array) is an essential component in the acquisition system, its adoption greatly decreases the acquisition size, however its design directly relates to the imaging quality. In this paper, a novel method of joint design of color filter arrays and demosaicing based on the spatial and multi-spectral correlation is proposed. The isotropic neighborhood is exploited for acquiring the spatial correlation, and the multi-spectral is utilized for the correlation cross channels. A convolutional network autoencoder is presented to design the color filter array patterns and demosaic, which comprehensively employs the spatial and multi-spectral correlation. Experiments shows that the method achieved high-quality image reconstructions, holding higher PSNR than the ones obtained by the state-the-art techniques on standard dataset.