Researchers from the Shenzhen Institute of Advanced Technology (SIAT) at the Chinese Academy of Sciences (CAS) and their collaborators have developed an efficient complex-valued attention mixer (ECA-Mixer) architecture for fast and accurate physics-informed color holographic reconstruction. 

The study was published in Optics and Lasers in Engineering on October 23. 

Color lens-free holographic imaging with an excellent field of view and spatial resolution has potential applications in medical pathology imaging. However, it faces a significant challenge in color holographic reconstruction through iterative optimization, hindering its practical application. 

While deep learning-based methods have gained increased interest in solving inverse imaging problems, they often neglect the diffraction wave transformation in complex space, leading to issues such as color aberration, long convergence times, and limited feature extraction abilities. 

To address these challenges, the researchers proposed a novel physics-informed color holographic reconstruction technology. Their method first applied a mixer architecture in complex space to enable the direct calculation of complex values with fewer parameters and shorter convergence times. Specifically, they introduced the use of 2D Haar wavelet and its inverse transform for encoding and decoding features, which were used to preserve high-frequency information. 

Simulated and experimental results demonstrated that the proposed method achieved state-of-the-art color reconstruction results in terms of computation time and image quality. Moreover, the proposed lightweight network is capable of imaging large-scale wide-field samples at high resolution in a matter of minutes. 

"Our work provides a new solution for the application of computational holographic imaging in biomedical microscopy," said Professor QIN Wenjian, the corresponding author of this paper. 

A demonstration of the large-scale wide-field color holographic reconstruction of Earthworm T.S by our proposed ECA-Mixer architecture. (Image by SIAT)

Media Contact:
ZHANG Xiaomin
Email:xm.zhang@siat.ac.cn