Search for a command to run...
Abstract Analog optical image processing provides a fast and energy-efficient computing paradigm capable of overcoming the high power consumption and latency limitations of conventional electronic chips in real-time, high-throughput image analysis. Recently, metasurfaces have emerged as a powerful platform for implementing diverse computing functions, including edge detection, convolution, and classification. To advance toward integrated optoelectronic image-processing units, it is crucial to achieve rapid switching between processing functions and enhance the flexibility of metasurface functional designs. However, existing image-processing metasurfaces typically suffer from slow tuning speeds and limited control over angular response tailoring. To address these challenges, we propose and theoretically investigate fast tunable metasurfaces for switchable image processing based on quasi-bound-state-in-continuum (qBIC). Numerical simulations demonstrate that femtosecond-laser pumping of silicon metasurfaces supporting qBIC modes in terahertz range enables microsecond-level fast switching between edge-detection and bright-field imaging functionalities. Furthermore, by combining the photonic tight-binding method with temporal coupled-mode theory, we establish the relationships between qBIC angular dispersion and the corresponding image processing performance. Finally, we explore qBIC metasurfaces with varying quality factors to analyze their switching and processing performances, providing quantitative insights into the design flexibility of image-processing functionalities. This work demonstrates a fast switching mechanism suitable for integrated optoelectronic processing units and offers a systematic theoretical framework for understanding and tailoring qBIC resonances in analog image processing.