Complex-Amplitude-Modulated Meta-Device for Optical Image Processing

Nowadays, convolutional neural networks (CNNs) have become a powerful tool in areas such as object recognition, and natural language processing (NLP). However, considering that electronic convolutional operation always contains million-level parameters and complex calculation process, it consumes a large number of computing resources and time. To overcome these limitations, we propose a design of complex-amplitude-modulated meta-device which could perform various functions of image processing. In this work, we demonstrate the excellent performance of two-dimensional edge detection and Gaussian filtering. The proposed convolutional system serves as a new optical computing hardware, and provides a new approach for CNNs, biological microscopy and near-infrared imaging.