A camera setup developed by Kobe University researchers is able to record 3D movies with a single pixel. Moreover, the technique can obtain images outside the visible spectrum and even through tissues. According to the researchers, the technique opens the door to holographic video microscopy. Traditionally, holograms have required a laser for recording, but more recently, techniques that can record holograms with ambient light or light emanating from a sample have been developed. There are two main techniques that can achieve this: “FINCH,” which uses a 2D image sensors that is fast enough to record movies, but is limited to visible light and an unobstructed view; and “OSH,” which uses a one-pixel sensor and can record through scattering media and with light outside the visible spectrum, but can only practically record images of motionless objects. A research effort led by assistant professor Yoneda Naru sought to create a holographic technique that combines the strengths of both techniques. To tackle the speed-limiting weak point of OSH, he and his team constructed a setup that uses a high-speed “digital micromirror device” to project onto the object the patterns that are required for recording the hologram. “This device operates at 22 kHz, whereas previously used devices have a refresh rate of 60 Hz. This is a speed difference that’s equivalent to the difference between an old person taking a relaxed stroll and a Japanese bullet train,” Yoneda said. The work demonstrates that the setup can not only record 3D images of moving objects, but the possibility of constructing a microscope capable of recording a holographic movie through a light-scattering object — more specifically, a mouse skull. The frame rate, however, is still fairly low, at just over one frame per second. Yoneda and his team showed in calculations that they could theoretically bring the frame rate up to 30 Hz., which is a standard screen frame rate. This would be achieved through a compression technique called “sparse sampling,” which works by not recording every portion of the picture all the time. “We expect this to be applied to minimally invasive, three-dimensional biological observation, because it can visualize objects moving behind a scattering medium. But there are still obstacles to overcome,” Yoneda said. “We need to increase the number of sampling points, and also the image quality. For that, we are now trying to optimize the patterns we project onto the samples and to use deep-learning algorithms for transforming the raw data into an image.” The research was published in Optics Express (www.doi.org/10.1364/OE.560998).
