MechResCon2023: Real-time Hologram Rendering from Interferometers is Made Possible by BitFlow Frame Grabber
Around 70 years after the invention of holographic imaging, this original wave-front reconstruction method has been enhanced due to improvements in image sensors and CPU power. Phase contrast imaging for samples in biological and biomedical studies, forensic identification, and MEMS vibration analysis, among other applications, has proven to be a practical method using the original optical reconstruction process, which is now carried out digitally and is known as Digital Holography Microscopy (DHM).
DHM typically calls for substantial post-processing computations and offline wave propagation, which can take a lot of time and be expensive. However, DHM may now be carried out at large input and output throughputs using ordinary components thanks to a novel system1 created at Institut Langevin, an academic research institute in Paris.
The Institut Langevin system makes use of the image capabilities of an Adimec Quartz 2MP CoaXPress camera and a Bitflow Cyton-CXP CoaXPress frame grabber. Rendering and processing are carried out using an Nvidia GTX Titan Xp card using Cuda 9. In addition to real-time hologram generation from interferograms using angular spectrum propagation and short-time Fourier transform, Holovibes, a flexible computational software for producing digital holograms, also offers buffered picture acquisition at high throughput.
At the Institut Langevin, real-time image rendering was experimentally shown using Mach-Zehnder interferometers and single-frequency lasers with wavelengths of 840, 532, and 780 nm. Images depicted the blood flow in mice's brains. Complex-valued holograms were created from an input stream of 1024 x 1024 (16-bit/pixel) digitised interferograms, which were captured by the BitFlow Cyton CXP at a throughput of 1.8 GB/sec. From these holograms, spectral components at about 100 Hz and 450 Hz were presented. The front and posterior segments of a human eye were computed and seen in real-time using the same setup using an input stream of 2000 frames per second (2GB/sec), 16-bit, 1024 by 512 pixel interferograms. Principal component analysis was used to demodulate the temporal signal of 32 consecutive hologram stacks at a rate of 60 hertz. Both tests were a resounding success.
This method can aid supervised machine learning in addition to offering real-time hologram rendering for various measurements and monitoring. It can efficiently produce enormous datasets of holographic reconstructions from digital interferograms, which might be used to train deep neural networks.
In his January 2018 conference paper, "Ultrahigh-throughput rendering of digital holograms," Michael Atlan of the Institut Langevin, French National Center for Scientific Research
URL of the source: bitflow.com
Comments
Post a Comment