Color filter arrays (CFA) have enjoyed immense popularity in single sensor imaging, including consumer, professional and scientific imaging. Owing in part to the technical breakthroughs that led to exceedingly high sensor resolution and high quality CFA interpolation algorithms, color image sensors now deliver high quality images. In this work, we investigate the plausibility of extending CFA sampling to enable multispectral imaging capabilities. In contrast to the prior work in this area that have been based on heuristics, we take a principled modeling approach based on a three dimensional Fourier transform (2D space, 1D spectrum) that reveal surprising degree of “spatial-spectral” structure. We propose a new spectral filter array (SFA) design aimed at maximizing the “recoverability” of the multispectral image signals. We provide concrete trade-offs between spatial and spectral resolution stemming from SFA sampling strategies.
Keigo Hirakawa, “Spectral Filter Array Design For Multispectral Image Recovery,” IAPR CCIW, 2011. [keynote]
Keigo Hirakawa, Paul M. Simon, “Spatial-Spectral Filter Array Approach To Multispectral Imaging,” under preparation. [bibtex][pdf]