A hyperspectral imaging camera that fits in your pocket

Prof. Min H. Kim’s team at the KAIST School of Computing has developed a new hyperspectral imaging method that enables hyperspectral imaging with only a single diffractive optical element. The method does not require any additional optics for snapshot hyperspectral imaging technology. This is the core technology that will make hyperspectral imaging available in conventional smartphone cameras.

Article | Spring 2020

In 1704, Isaac Newton discovered that light is a mixture of differently refrangible rays. What if we could capture the rich spectra of different wavelengths for every pixel in a digital photograph? Spectral information is crucial for various imaging applications in remote sensing, military defense, cultural heritage digitalization, agriculture, mineralogy, and computational photography. These applications require snapshot capability and a small form factor of the imaging systems.

Existing hyperspectral imaging systems have captured rich spectral information of real-world scenes. However, they use spatio-spectral pushbroom scanning, which makes it difficult to capture dynamic scenes. Their form factors are significantly larger than those of conventional digital cameras, so they cannot fit into our pocket. Many users would benefit from more compact and portable imaging systems.

To address this need, Prof. Min H. Kim’s team at the KAIST School of Computing has developed a new hyperspectral imaging method that enables hyperspectral imaging with only a single diffractive optical element. The method does not require any additional optics for snapshot hyperspectral imaging technology. Prof. Kim noted, “This is the core technology of snapshot hyperspectral imaging that will make hyperspectral imaging available in our conventional smartphone cameras. Using only a single diffractive optical element, we can make a hyperspectral camera as compact as possible so that the spectral information of the objects can be captured by simply taking a single photograph from a camera.”

The research team demonstrated their camera system in a series of examples in their paper, “Compact Snapshot Hyperspectral Imaging with Diffracted Rotation.” The novel design of the diffractive optical element and the artificial intelligence-based algorithm required no additional optical element, successfully captured rich spectral information from a single shot photograph and reproduced consistent results. Examples showcased in the work included various sets of ordinary scenes with different spectra.

Prof. Kim and his collaborators, Prof. Wolfgang Heidrich of King Abdullah University of Science and Technology (KAUST) in Saudi Arabia, and KAIST PhD students, Daniel S. Jeon and Seung-Hwan Baek, published the new research work in ACM Transactions on Graphics (TOG) and presented it at ACM SIGGRAPH 2019 in Los Angeles, U.S.A. In future work, they plan to further extend the capability of the method or to make it more compact for use in mobile phones (http://vclab.kaist.ac.kr/siggraph2019/).