Reconfigurable nonlinear nonlocal flat optics for analog image classification and transformation
Year: 2026
Authors: Goh H., De Angelis C., Kiani M., Deriy I., Kim S., Alsch A.
Autors Affiliation: CUNY, Adv Sci Res Ctr, Photon Initiat, New York, NY 10031 USA; Seoul Natl Univ, Dept Civil & Environm Engn, Seoul 08826, South Korea; Seoul Natl Univ, Inst Construct & Environm Engn, Seoul 08826, South Korea; Univ Brescia, CNIT, Via Branze 38, I-25123 Brescia, Italy; Univ Brescia, Dept Informat Engn, Via Branze 38, I-25123 Brescia, Italy; CNR, Ist Nazl Ott, Via Branze 45, I-25123 Brescia, Italy; CUNY, Grad Ctr, Phys Program, New York, NY 10016 USA; RTX BBN Technol, Cambridge, MA 02138 USA.
Abstract: Wave-based analog image processing metasurfaces have emerged as a powerful real-time and energy-efficient platform for the execution of repetitive and data-intensive image operations within an ultrathin platform, with Prof. Engheta being one of the early seminal contributors to this rapidly growing field of research. However, to date, existing analog optical processors have been limited to low-level linear canonical tasks, such as edge detection and denoising, and they are typically difficult to reconfigure for different operations. In this work, we discuss a flat optical platform for analog image classification and transformation by exploiting nonlinear, nonlocal wave-matter interactions. Specifically, we leverage pointwise multiplication enabled by the second-order nonlinear susceptibility to quantify image similarity for classification and to enable coordinate transformations, where the image operation can be dynamically reprogrammed by controlling the incident wave field. We further demonstrate parallel processing of multiple computations using phase-configured input images and explore the metasurface potential as a computing reservoir. This approach offers substantial improvements in performance and efficiency for real-time image processing applications. (c) 2026 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement
Journal/Review: OPTICAL MATERIALS EXPRESS
Volume: 16 (10) Pages from: 3333 to: 3341
More Information: The authors acknowledge fruitful discussions with Kristina Friziuk. This work was supported by the Air Force Office of Scientific Research MURI program and by the Simons Foundation. H. G. was partially supported by National Research Foundation of Korea (NRF) grant (No. 0583-20250056).KeyWords: Analog image processing; Energy efficient; Flat optics; Image transformations; Images classification; Metasurface; Nonlocal; Real- time; Reconfigurable; Time-efficientDOI: 10.1364/OME.611457
