近日，美国加州大学的Aydogan Ozcan及其研究团队取得一项新进展。经过不懈努力，他们提出用于单向图像放大和缩小的金字塔衍射光学网络。相关研究成果已于2024年7月31日在国际知名学术期刊《光：科学与应用》上发表。

该研究团队提出了一种金字塔结构的衍射光网络设计（命名为P-D²NN），该设计专门针对单向图像放大和缩小进行了优化。在此设计中，衍射层按照图像放大或缩小的方向呈金字塔状排列。这种P-D²NN设计能在单一方向上创建高保真度的放大或缩小图像，同时抑制相反方向的图像形成，从而在光学处理器体积内使用更少的衍射自由度，实现所需的单向成像操作。另外，尽管P-D²NN设计是使用单一波长进行训练的，但它能在大范围的照明波长内保持其单向图像放大/缩小的功能。

研究人员还设计了一种波长复用的P-D²NN，其中的单向放大器和单向消音器能在两个不同的照明波长下，同时在相反的方向上工作。此外，研究人员证明，通过级联多个单向P-D²NN模块，可以获得更高的放大系数。P-D²NN架构的有效性也在太赫兹照明下得到了实验验证，实验结果与数值模拟高度吻合。P-D²NN为设计针对特定任务的视觉处理器提供了一种受物理学启发的新策略。

据悉，衍射深度神经网络由连续传输层组成，利用监督深度学习进行优化，以全光学方式实现输入和输出视场之间的各种计算任务。

附：英文原文

Title: Pyramid diffractive optical networks for unidirectional image magnification and demagnification

Author: Bai, Bijie, Yang, Xilin, Gan, Tianyi, Li, Jingxi, Mengu, Deniz, Jarrahi, Mona, Ozcan, Aydogan

Issue&Volume: 2024-07-31

Abstract: Diffractive deep neural networks (D²NNs) are composed of successive transmissive layers optimized using supervised deep learning to all-optically implement various computational tasks between an input and output field-of-view. Here, we present a pyramid-structured diffractive optical network design (which we term P-D²NN), optimized specifically for unidirectional image magnification and demagnification. In this design, the diffractive layers are pyramidally scaled in alignment with the direction of the image magnification or demagnification. This P-D²NN design creates high-fidelity magnified or demagnified images in only one direction, while inhibiting the image formation in the opposite direction—achieving the desired unidirectional imaging operation using a much smaller number of diffractive degrees of freedom within the optical processor volume. Furthermore, the P-D²NN design maintains its unidirectional image magnification/demagnification functionality across a large band of illumination wavelengths despite being trained with a single wavelength. We also designed a wavelength-multiplexed P-D²NN, where a unidirectional magnifier and a unidirectional demagnifier operate simultaneously in opposite directions, at two distinct illumination wavelengths. Furthermore, we demonstrate that by cascading multiple unidirectional P-D²NN modules, we can achieve higher magnification factors. The efficacy of the P-D²NN architecture was also validated experimentally using terahertz illumination, successfully matching our numerical simulations. P-D²NN offers a physics-inspired strategy for designing task-specific visual processors.

DOI: 10.1038/s41377-024-01543-w

Source: https://www.nature.com/articles/s41377-024-01543-w