Engineers at the Massachusetts Institute of Technology have developed a new technique to reveal objects in darkened images using a physics-based algorithm and a trained neural network.
MIT researchers have published their work in Physical Review Letters today showcasing how they reconstructed transparent objects from photos of those objects, taken in a nearly pitch-black environment.
Spotting an imperfection in glass such as a micro-crack or crease is difficult to do even in perfect lighting conditions. Asking a computer to decipher and identify a crack in a transparent object from a photo taken in low lighting conditions is no easy task.
The MIT researchers trained a neural network to recognise over 10,000 transparent glass-like etchings, from low-level photos taken of them. Once they introduced a new grainy image that the computer had not seen before it was able to reconstruct the darkened image using its training.
Biomedical Sector
The practical use of this technology can be found in the biomedical field as MIT researchers have demonstrated that a deep neural network can be used to illuminate transparent features of cells and tissue samples.
“In the lab, if you blast biological cells with light, you burn them, and there is nothing left to image,” commented George Barbastathis professor of mechanical engineering at MIT in an MIT research blog .
“When it comes to X-ray imaging, if you expose a patient to X-rays, you increase the danger they may get cancer. What we’re doing here is, you can get the same image quality, but with a lower exposure to the patient. And in biology, you can reduce the damage to biological specimens when you want to sample them.”
See Also: Research Teams Use Summit Supercomputer to Win Gordon Bell Prize
The researchers constructed an experiment where they pointed a camera at a small aluminum frame that contained a phases spatial light modulator, an instrument that recreates the same optical effect an etched slide would. They took images of each of the 10,000 dataset images in near dark condition. These images resembled what you would see if you had static on a TV screen.
The neural network was able to reconstruct these grainy images to an extent that details and objects can be recognised in them.
The papers co-author Alexandre Goy commented that: “We have shown that deep learning can reveal invisible objects in the dark. This result is of practical importance for medical imaging to lower the exposure of the patient to harmful radiation, and for astronomical imaging.”
The post Researchers Train Neural Network to Reveal Objects in Grainy Images appeared first on Computer Business Review.