Single-pixel complex-amplitude imaging based on untrained complex-valued convolutional neural network.

Area of Science:

• Optics and Photonics
• Computational Imaging
• Artificial Intelligence in Imaging

Background:

• Single-pixel imaging (SPI) is a rapidly developing field for complex-amplitude imaging.
• Current SPI methods require extensive data acquisition and computational resources, leading to lengthy imaging processes.
• High-quality reconstruction in complex-amplitude imaging remains a challenge.

Purpose of the Study:

• To propose a novel single-pixel complex-amplitude imaging (SCI) scheme for efficient and high-quality image reconstruction.
• To develop a reconstruction method that bypasses the need for pre-trained labeled data.
• To establish a new framework for SCI applications.

Main Methods:

• A novel single-pixel complex-amplitude imaging (SCI) scheme utilizing a complex-valued convolutional neural network (CNN) for image reconstruction.
• The CNN is randomly initialized and operates solely under the constraints of the physical imaging model, eliminating the need for pre-training on labeled datasets.
• The proposed method focuses on optimizing the balance between imaging efficiency and reconstruction quality.

Main Results:

• Simulation and experimental results demonstrate the effectiveness and feasibility of the proposed SCI scheme.
• The method successfully reconstructs high-quality complex-amplitude images with significantly reduced computational time compared to traditional methods.
• A favorable trade-off between imaging efficiency and image quality was achieved.

Conclusions:

• The developed SCI scheme offers a new and effective framework for reconstructing complex-amplitude images.
• The proposed method significantly improves the efficiency of single-pixel imaging without compromising image quality.
• This work paves the way for broader practical applications of single-pixel complex-amplitude imaging.