The world of optics is indeed fascinating, with several developments rolling out now and then. These developments, in the form of inventions and surprising discoveries, have major ripple effects on the world as we see it.
An understanding of optics, a study of the behaviors and properties of light is central to imaging. Developments in the technologies related to visible and invisible light, such as lenses and fiber optics, also impact photography.
Here is a look into a recent development in the world of imaging caused by combining optics with machine learning.
Lenses, optics and imagery
Conventionally, several layers of glass lenses are stacked to create the refined imaging that we get through cameras, which means that cameras cannot be resized to become smaller beyond a certain limit. In this age of advancement, we are constantly looking to improve upon existing technology. Why? Because bulkier gear is not as attractive anymore.
In the search for smaller optics, scientists created meta surface optics that use sub-wavelength antennas to reproduce images. Meta lenses are very thin lenses that use hundreds of thousands of nano-antennas that capture and reflect light in terms of nanometers.
The only problem researchers found with this technology was that the quality of the image was very poor. This was because there were aberrations at larger apertures and lower f-stops.
In optics, lens aberration is a property that causes an image formed by a lens to be blurry. This occurs because the light rays going through the lens do not converge at one point. Instead, they spread out and focus on different points, leading to an unclear image.
To combat these two issues, a team of scientists from Princeton and Washington University have created neural nano-optic imagers using AI and machine learning.
What are neural nano optics?
Simply put, neural nano-optics are very small and thin optical metasurfaces powered by machine learning. The ultra-small imager is as small as a grain of salt! This makes it the smallest possible camera with an f-stop value of f/2 and an extra-large aperture of 0.5mm!
Neural nano-optics can reproduce images jointly with deep learning algorithms reconstructing these images on the other end.
The team of scientists who created this belief that the image quality produced by these thin lenses are “on par” with the quality produced by commercial compound lenses that are over 550,000 times bigger. (Tseng E. et al, 2021)
How does it work?
The principle behind the creation and function of nano-optics is the use of optical meta surfaces with AI techniques. Meta surfaces are specially designed sheet material comprising sub-wavelength antennas that manipulate the behavior of electromagnetic waves.
Optical meta surfaces manipulate light by first capturing the photons, then re-emitting them after “bending” the electromagnetic waves. This is done by using the sub-wavelength cylindrical posts on a small ultra-thin square surface. These cylindrical posts act as antennas to capture electromagnetic waves and direct them in a refined manner.
After this, the optical meta surface sends signals to a computer powered by deep learning to produce a final image. The result is accurate high-quality color images created using deep learning algorithms that interpret the signals sent by the meta surface.
According to Tseng and the other researchers, their nano-camera powered by neural nano-optics is the first meta-optics imager to achieve high-quality wide-field of view color images. There is a significant improvement in the resolution and image quality, with reduced aberrations. Its quality is aeons ahead of meta surface imagers.
This technology closes the gap between the size, potential and performance of the imager by producing high-quality images at a much more efficient rate. It is crucial for the development of sophisticated new-age smartphones that see innovations every other quarter. Thin lens imaging using neural nano-optics can replace the bulky compound optics present in telescopes, microscopes, DSLR cameras and AR/VR gear.
Applications of neural nano-optics
Any discovery or innovation in the field of optics massively impacts other fields that depend on vision and being able to capture what we see around; this impacts the world of photography.
Here are a few key applications of neural nano-optics
- Smartphone cameras
One of the biggest elements of a smartphone that developers and mobile giants take pride in and spend a lot of their revenue refining is the camera.
Smartphone cameras are, for many people, the single biggest factor that determines what smartphone brand and model they will choose. Developers flaunt the inclusion of several cameras of varying resolution, in a bid to attract as many customers as possible. This also means including cameras with several layers of lenses that create an overall bulkier feel. This isn’t exactly very attractive nor is it feasible today.
The introduction of neural nano-optics would then prove to single-handedly elevate the quality of imaging without the need for several cameras. In the next few years, nano-optics will completely shift the paradigm and replace the present cameras in smartphones, significantly impacting how we photograph objects and view them.
- Medical imaging
The most obvious application is in the field of medicine, which will greatly benefit from nano-sized ultra-thin cameras that can be used for capturing images of internal organs.
Obtaining clearer images of what is going on inside our bodies can also aid in pinpointing the exact causes and help doctors decide on the best approach for treatment. If implemented en masse, neural nano-optics can have groundbreaking effects and revolutionize our approach to these fields.
Also read: AI & Deep learning based image denoising solutions | Insights - Tooliqa
Tooliqa specializes in AI, Computer Vision and Deep Technology to help businesses simplify and automate their processes with our strong team of experts across various domains.
Want to know more on how AI can result in business process improvement? Let our experts guide you.
Reach out to us at firstname.lastname@example.org.