MEPhI graduate, Candidate of Physical and Mathematical Sciences Pavel Cheremkhin knows everything about digital holograms. Or almost everything. The work of the employees of the LaPlaz Laboratory of Photonics and Optical Information Processing, associate professor Pavel Cheremkhin, master student Andrey Svistunov, postgraduate student Dmitry Rymov and professor Rostislav Starikov "Reconstruction of a digital hologram using a branched neural network" was recently published in the scientific journal Applied Sciences. The researchers proposed a new type of neural network image recovery from holograms - faster and more accurate. Their innovative developments in this area are already being applied in practice.
The principle of holography was formulated in 1947 by the Hungarian-British scientist Dennis Gabor, but it was only possible to fully implement it after the creation of the laser. The laser radiation is divided into two beams: reference and subject, which illuminates the holographed object. A hologram is a registered picture from interconnected reference and object beams. The world's first three-dimensional hologram was recorded in the Soviet Union by the optical physicist Yuri Denisyuk. It happened back in 1962.
Later it turned out that the holograms registered in this way can be restored in white light (electromagnetic radiation in the visible range), so they are widely used in everyday life. For example, recently a holographic zoo opened in Australia, where all animals, including long-extinct ones, are replaced by their digital holograms. Of course, holograms not only entertain, but also help make our lives better and more comfortable, solve scientific, medical, and environmental problems. At first, holograms were recorded on photographic plates, but with the development of digital technology, it became possible to save them as computer files. This is very simple to explain. In fact, creating a high-quality hologram is a complex and time-consuming task.
A hologram is an optical image of an object that stores information not only about its shape, but also about its structural features. Scenes can be reconstructed using computer methods of calculation or optical display of holograms on special microdisplays – spatiotemporal light modulators. At the same time, it is important that with the help of holograms it is possible to restore information about objects as accurately as possible, despite the various negative factors present during registration. Pavel Cheremkhin is working on this task.
“Digital holography is a technology that allows recording and processing data on the structure and characteristics of objects in two and three dimensions. This technology has become possible due to the development of optical-digital methods for recording images and their computer processing. It is enough to take a picture of the resulting picture in order to restore and supplement information about objects, as well as to find out their important characteristics. For example, by examining any material, we can obtain information about the presence of inclusions, its density or the slightest roughness on its surface. Thanks to the possibility of high-speed recording of digital holograms, it is also easy to track the dynamics of changes in the characteristics of an object,” says Pavel Cheremkhin.
The technology can be used in 3D monitoring systems, in nano- and microelectronics, in the study of macro- and micro-objects, such as microplastics. Say, in order to explore the "garbage island" in the Pacific Ocean. The Great Pacific Garbage Patch was formed from plastic waste and is expanding rapidly, posing a massive threat to the ocean's ecosystem. Thanks to digital holography, you can accurately find out what is the concentration of plastic in each specific place of the “island”, as well as determine its density, particle size, and track the dynamics of changes. Which, in turn, will contribute to the development of the right strategy to cleanse the ocean. All you need is a laser machine and digital cameras that will record light reflected from water and microplastic particles. The resulting hologram will allow you to see how the plastic interacted with the environment and analyze all reactions in real time. True, in complex cases, which include the same island of microplastics in the ocean, so-called noises (random changes in brightness or color information) are superimposed on the image of the object, so the hologram may turn out to be insufficiently clear, weakly standing out against the general background.
The task of researchers from the MEPhI was to restore high-quality images of digital holograms for complex cases, when there are many objects, they are distributed throughout the volume and the noise level is increased. The advantage of the method proposed by the scientists of our university is the use of neural networks to speed up calculations and reduce the noise level.
“To get rid of noise, registration of a whole set of holograms is usually used. But this reduces the speed of research. We proposed a new type of neural network image recovery from holograms and were able to teach the neural network to immediately restore many 2D sections of a 3D scene. Recovery calculation takes hundredths or even thousandths of a second! And the images are obtained free of noise and give a fairly accurate position of each of the micro-objects. You can estimate their number, size, density. If we record a holographic video, then we can watch how objects change and track the dynamics in real time,” explains Pavel Cheremkhin.
Digital holograms have many applications and in the near future, the scientist believes, will be used even more often. In medicine, they can be used to perform surgical operations and clinical tests (holography allows, for example, to determine not only the number of red blood cells in the blood, but also their size and possible deformations, and with very high accuracy). The method can be useful in the analysis of dynamic and fast processes, where it is necessary to quickly determine the location of particles in a dispersed medium. Digital holography can make tutorials more vivid and visual, more accurate architectural designs or gadget assembly. And, of course, it will be nice to have a holographic video call with friends and see how their holographic images appear right in front of us.
More details about the restoration of 3D images using digital holography and the proposed HoloForkNet method can be found in the work “Reconstruction of digital holograms using a branched neural network of a neural network”, which was performed under a grant from the Russian Science Foundation. Research on this topic is also supported by the Priority 2030 program.
