Shanghai Optics and Fine Mechanics Institute has made progress in the new holographic imaging technology of frequency domain direct sampling

Recently, a research team from the Aerospace Laser Technology and Systems Department of the Shanghai Institute of Optics and Fine Mechanics, Chinese Academy of Sciences, proposed a new holographic imaging technology using frequency domain direct sampling. The relevant results were published in Optics Letters under the title of "Fourier inspired single pixel holography".

Digital holography is a technique that uses interference to record information about the optical field. Among them, off-axis digital holography is widely used in imaging, measurement, display, storage and other fields due to its ability to eliminate the influence of twin images in principle. Traditional off-axis digital holography uses an array detector to record holograms, and then selects the frequency spectrum related to the target light field. This imaging method records both the target image and redundant zero order images and twin images. In addition, due to the difficulty in preparing wide spectrum, high sensitivity, and high spatiotemporal resolution area array detectors, off-axis digital holography technology is also difficult to apply under extreme conditions such as special bands and low light.

This study is based on the mechanism of off-axis holography to separate redundant information in the spectral domain, combined with the characteristic of Fourier single pixel imaging technology that can obtain object spectra on demand. The hologram is encoded using a specific frequency stripe pattern, and the Fourier spectrum of the target light field is directly sampled by a high-sensitivity single pixel detector. Finally, the target light field information is obtained through inverse Fourier transform (Figure 1a, b). In addition, the research team intelligently designed encoding patterns and image enhancement models based on self coding architecture (Figure 1c), and used transfer learning techniques to reduce the required experimental data volume (Figure 1d). In the end, the research team successfully achieved imaging of phase type objects at a maximum sampling rate of 7.5% (Figure 2). This study combines holography, correlation imaging, and artificial intelligence organically, providing new ideas for efficient phase detection in special bands and low light conditions. It is expected to be applied in fields such as scattering imaging and low light imaging.

Figure 1. (a) Off axis hologram generation process; (b) The process of modulating off-axis holograms using learned mask selection patterns and reconstructing objects; (c) The joint optimization network structure used for pre training; (d) Fine tuning process using experimental data.

Figure 2. (a) Experimental setup diagram; (b) The original experimental results at different sampling rates and the experimental results processed by neural networks; (c) Phase truth, phase comparison of network input and network output objects at a sampling rate of 7.5%.

Source: opticsky