Two-dimensional optical differentiator for broadband edge detection based on dielectric metasurface

Image edge processing has widespread adoption in a variety of scientific and industrial scenarios. To date, implementations of image edge processing have mostly been done electronically, but there are still difficulties to achieve real-time, high-throughput, and low power consumption image edge processing. The advantages of optical analog computing include low power consumption, fast transmission speed, and high parallel processing capability, and optical analog differentiators make this process possible. However, the proposed analog differentiators can hardly meet the requirements of broadband, polarization insensitive, high contrast, and high efficiency at the same time. Moreover, they are limited to one-dimensional differentiation or work in reflection mode. To be better compatible with two-dimensional image processing or image recognition systems, two-dimensional optical differentiators that integrate the above advantages are urgently needed. In this Letter, a two-dimensional analog optical differentiator with edge detection operating in transmission mode is proposed. It can cover the visible band, is polarization uncorrelated, and has a resolution that reaches 1.7 μm. The efficiency of the metasurface is higher than 88%.