Metasurface-empowered snapshot hyperspectral imaging with convex/deep (CODE) small-data learning theory

Abstract

Hyperspectral imaging is vital for material identification but traditional systems are bulky, hindering the development of compact systems. While previous metasurfaces address volume issues, the requirements of complicated fabrication processes and significant footprint still limit their applications. This work reports a compact snapshot hyperspectral imager by incorporating the meta-optics with a small-data convex/deep (CODE) deep learning theory. Our snapshot hyperspectral imager comprises only one single multi-wavelength metasurface chip working in the visible window (500-650 nm), significantly reducing the device area. To demonstrate the high performance of our hyperspectral imager, a 4-band multispectral imaging dataset is used as the input. Through the CODE-driven imaging system, it efficiently generates an 18-band hyperspectral data cube with high fidelity using only 18 training data points. We expect the elegant integration of multi-resonant metasurfaces with small-data learning theory will enable low-profile advanced instruments for fundamental science studies and real-world applications.

Electrical-Driven Dynamic Augmented Reality by On-Chip Vectorial Meta-Display

“…The optical extraction efficiency can be further improved by utilizing low-loss and high-index dielectric materials (such as titanium dioxide, TiO 2 ), appropriately increasing the metasurface area, or applying more precise fabrication processing. In addition, some artificial intelligence methodologies for meta-optics can be utilized, including inverse design, deep learning, , and other approaches, , to further optimize the efficiency.…”

Freeform metasurface color router for deep submicron pixel image sensors

“…Further research could include designs more resistant to obliquely incident light, hard fabrication constraints, or pattern generation algorithms suitable for mass production. Recent high-performance metasurface devices have been designed using deep learning algorithms or a combination of different optimization algorithms ( 24 , 31 , 46 , 51 ). The combination of deep learning and optimization algorithms used here would potentially find a solution closer to the global minimum.…”

“…The progress on the metasurface optics has emerged as an influential platform in the field of diffractive optics, with the benefits of compactness and their high control capacities to modulate electromagnetic waves ( 3 – 7 ). These advantageous features open up avenues for expanding functionalities of optical elements and miniaturizing optical systems, as demonstrated in various applications, such as metalenses ( 8 – 17 ), holograms ( 18 – 20 ), deflector ( 3 , 21 – 24 ), color filters ( 25 ), optical computation ( 26 ), phase imaging ( 27 , 28 ), polarization imaging ( 29 , 30 ), hyperspectral imaging ( 31 ), and active devices ( 32 , 33 ). The use of metasurface and volumetric meta-optics has allowed for the development of color router devices that can split and focus the incident visible light to a designated photodiode instead of simply filtering out colors ( 34 – 43 ).…”

Metasurface-based computational imaging: a review

“…Through combining convex optimization and deep learning, 18 channels of hyperspectral image are retrieved from the captured image with a small amount of data for training the neural network. 54 In summary, by fully harnessing the inherent redundancy in hyperspectral images and integrating advanced computational imaging techniques, there is optimistic anticipation that highquality hyperspectral imaging retains essential spectral nuances that can be promisingly realized.…”