Abstract-Resolution Enhancement in Terahertz Imaging via Deconvolution

Wei Ning, Feng Qi, Zhaoyang Liu,  Yelong Wang, Hongming Wu, Jinkuan Wang

https://ieeexplore.ieee.org/document/8717628

Nowadays, terahertz (THz) imaging is quite promising in scientific and biomedical applications. However, the much longer wavelength degrades the resolution of obtained images compared with classical imaging at optical frequencies. Imaging is considered to be a convolution process between the object function and point spread function (PSF) of the system. Thus, how to describe the object accurately from a blurred image is an inverse problem. In this paper, we demonstrate how well such a method can function. We would illustrate how to make current facilities overcome the physical limitations in imaging, which can image at the sub-spot level. Periodical stripes with a period of 0.8 mm are resolved at 0.3 THz. The resolution is 0.32 times the physical size of the focused beam. Moreover, this method is extended to the mathematical operations in a complex domain, which is distinct from previous image studies via visible/IR light. Physically, it means that we essentially deal with phase by making use of the wave nature of THz light. For the first time, it is demonstrated that the reconstructed phase can also describe the shape of the object by applying such an approach. In addition, we have successfully measured the diameter of human hair with a relative error of less than 10% at 3 THz. As an example, we did an experiment on tissues of colorectal cancer and a better image is obtained, which would be helpful for medical diagnosis. We believe that such an approach is able to improve imaging performance at THz frequencies and many applications will gain from it.

Terahertz Imaging Deconvolution Method Proposed to Enhance Resolution

THz imaging of Human hair. From left: photo of hair sample, THz image before enhancement, THz image after enhancement. (Image by QI Feng)  

http://english.cas.cn/newsroom/research_news/201906/t20190627_212170.shtml

QI Feng and his team at Terahertz Imaging Laboratory, Shenyang Institute of Automation of the Chinese Academy of Sciences, demonstrated a terahertz (THz) imaging deconvolution method which is able to image at the sub-spot level to achieve high-resolution THz images.

Their finding, published in IEEE Access, showed that phase information of THz waves is fully utilized by making use of the wave nature of THz light. The study demonstrated that the reconstructed phase can also describe the shape of the object by applying such an approach.

THz imaging is quite promising in scientific and biomedical applications. Current spatial resolution of THz imaging systems is not perfect due to its much longer wavelength compared to visible light. Many scientists have been working on near-field imaging systems to achieve a high resolution image. However, in this system objects have to be placed extremely close to the sensors, which is difficult to achieve in practical applications.

How to extract information beyond the physical limit of the beam size is important. Generally, imaging is considered to be a convolution process and the point spread function essentially blurs the figure. THz light, as a kind of electromagnetic waves, has both amplitude and phase information. Optical image processing algorithms are based on geometrical optics and only intensities are considered.

To evaluate the spatial resolution quantitatively, the researchers in this study fabricated a periodic structure sample with three 0.8 mm bar patterns. This sample can be resolved well at 0.3 THz, and the resolution is 0.32 times of the physical size of the focused beam.

In addition, they successfully measured the diameter of human hair with a relative error of less than 10% at 3 THz.

This study demonstrates a promising method that will be helpful for providing clearer imaging for scientific and biomedical applications.