Mask-based Ultra-thin Lensless Camera Design and Production Technology
Presented Possibility of Commercialization of Ultra-thin Camera with Arbitrary Point Spread Function Through Phase Mask Design and Single Exposure Production Technology
Published in ‘Optica (IF 10.644)’, the top academic journal in the field of optics, on January 8th
The research team of Professor Seung Ah Lee, from the Department of Electrical and Electronic Engineering at the College of Engineering, developed a technology that mass-produces new optical elements, a world first and the core of lensless camera technology, through collaborative research with the research team of Professor Wook Park, from the Department of Electronic Engineering at Kyung Hee University.
The research team produced a camera five times thinner than existing micro-miniature cameras. They demonstrated applications such as image capture as well as the implementation of optical information processing and presented the possibility of commercializing original lensless camera technology for the first time. In particular, this approach has a cost efficiency significantly higher than that of existing nano-processing methods, as it enables the design of various phase masks using a phase recovery algorithm and can quickly produce them with a single exposure. Through this and general photography, the commercialization of lensless camera technology using computational imaging techniques for various purposes, such as fingerprint and face recognition, can be expected.
As the use of cameras has expanded from simple photography to object recognition, biometric recognition, and information acquisition with the development of image sensors and artificial intelligence-based image processing technology, the demand for the miniaturization of camera modules is increasing. However, there are physical limitations to reducing the overall size and cost of camera modules while the pixel resolution of cameras continuously increases.
Accordingly, a lensless imaging technique that replaces a lens with a thin light modulation mask is attracting attention. Lensless imaging is a method of measurement by placing, instead of a lens, a thin mask in front of the sensor that modulates light. The thickness and focal length of the lens module are collectively reduced, thereby making it possible to produce ultra-thin cameras that exceed the physical limitations of existing cameras. Unlike conventional cameras in which one point of an object is matched to one point of the image sensor, a point of an object is matched in a two-dimensional pattern by the image sensor in a lensless camera, and the scene of the object is restored by applying an additional calculation algorithm to the measured value of the image sensor.
The phase mask structure is the most crucial factor determining the performance of the lensless camera. The surface of the phase mask has a fine curved structure; accordingly, a two-dimensional wave pattern is formed on the sensor surface when light from a point light source passes through the phase mask. This is called the Point Spread Function (PSF). As a two-dimensional PSF can be formed, a technology that can easily produce more diverse phase masks with a higher degree of design freedom is required.
The structure of the phase mask is calculated by implementing an optimization problem-based algorithm such that an arbitrary PSF pattern can be formed at a specific focal length. The phase mask designed in this manner is produced by a maskless grayscale lithography system. The produced system utilizes a Digital Micromirror Device (DMD) as a dynamic mask, irradiates a two-dimensional UV light pattern to a photocurable resin, and produces phase masks of various shapes with a single exposure. Since the DMD is used as a dynamic mask, phase masks with various structures can be instantly designed and produced with a single exposure. The process is straightforward and affords significant advantages in terms of production speed and cost over conventional nano-processing. Moreover, this technique produces a phase mask on an image sensor, and it is possible to combine it with the image sensor process.
The research team used this technology to produce phase masks with various PSFs. Developing an ultra-thin camera made it possible to image both the scene on display and objects in real life. In addition, this study presents a deep learning-based image restoration network for real-time restoration and restoration quality improvement. It demonstrates the transfer of a learned network using repeatability of production. When deep learning-based restoration is used, restoration results can be produced within milliseconds. Even with a camera that is not used for learning, images with much improved quality can be obtained compared to existing reconstruction algorithms.
Furthermore, the designed lensless camera can be used for optical information processing using the PSF of the phase mask as a kernel. In a lensless camera, the form of convolution of a two-dimensional PSF with an object is measured by an image sensor. Using this, it is possible to calculate whether a specific pattern matches an object optically. After producing a phase mask, which has some characteristics of the fingerprint as PSF, fingerprint matching can be confirmed with only the original data taken by the camera when the fingerprint is shown on display.
Professor Seung Ah Lee conveyed the significance of the study in the following words: “Studies on various smart imaging applications using lensless cameras have been reported recently. Through this technology, an ultra-thin lensless camera with an arbitrary transfer function could be commercialized. Accordingly, the development of ultra-small and low-cost smart camera modules optimized for various purposes such as object classification, recognition, and detection based on optical imaging is expected.”
This research was conducted by Kyung-Cheol Lee, a researcher at Yonsei University, and JunghyunBae, a researcher at Kyung Hee University, as the leading authors. This is an IT convergence study in which professor Seung Ah Lee and professor Wook Park of Kyung Hee University participated as corresponding authors. On January 8, it was published in ‘Optica (IF 10.644),’ the top academic journal in the field of optics. The research was conducted with support from the National Research Foundation of Korea (NRF-2021R1C1C101290012) and the KIST Support Program (2021R1C1C101290012).
Professor Jong-Hyun Ahn
Professor Seong Chan Jun
Professor Donghyun Kim