This article was originally included on Popular photography.
Every photographer knows this truth: when it comes to cameras and equipment, things can get huge – and fast. Especially in the lens department. Although mirrorless are first advertised as the smaller and lighter alternative to large DSLRs (and somehow they still keep that promise), it’s hard to ignore the fact that these once promising lightweight systems can quickly sunk in the glass. And while lenses have been part of the photographic process from the beginning, innovative research by a group of scientists from the Tokyo Institute of Technology could lead to new lens-free imaging systems and ultimately a lens-free future.
Challenges of a camera without lenses
The concept of a camera without lenses is not entirely new. However, so far the technology has not yet yielded useful results, and the computational time required to display the images is too lethargic for practical use.
Current technology requires solving a problem with prominent optimization, and iterative calculations (which means that many results can be returned) make it a long process. But Tokyo Tech’s proposal has the potential to make significant improvements in the process.
“Deep learning can help avoid the limitations of model-based decoding, as it can instead learn the model and decode the image through a non-iterative direct process,” the team said. explained on Phys.org. “Existing deep learning methods for lens-free images that use a convolutional neural network (CNN) cannot produce high-quality images. They are ineffective because CNN processes the image based on the connections of neighboring “local” pixels, while lensless optics convert local scene information into overlapping “global” information about all pixels of the image sensor through a property called “multiplexing”. . “”
Researchers at Tokyo Tech are proposing a new technology
The Tokyo Tech team proposed a new method for image reconstruction using a mathematical algorithm. The hardware consists of a mask and an image sensor, the mask encoding the incident light (light that falls on the object), which then throws models on the sensor.
Looking at these models, the human eye could not decipher any usable information – everything is presented as incomprehensible stains, like the trace left when you throw spaghetti on the wall or go to Jackson Pollock on canvas. However, the optical algorithm is able to decode the information, which gives quite accurate results.
“Without the limitations of the lens, a lensless camera can be ultra-miniature, which can allow new applications that are beyond our imagination,” said Prof. Masahiro Yamaguchi of Tokyo Tech.
How it works
The team is a pioneer in a technology called Vision Transformer (ViT), which can learn image characteristics in what they describe as a “hierarchical” way, thus avoiding traditional CNN processing. This means direct reconstruction without the necessary iterative calculations and reduction of approximation errors.
When CNN relies on local pixel information, ViT can use global functions in an image. According to experiments, the process is viable and the proposed camera creates high-quality images with processing time, which would allow real-time capture.
“The ultimate goal of a camera without lenses is to be tiny but still powerful. We are excited to be a leader in this new direction for next-generation imaging and sensor solutions, ”said the study’s lead author, Mr. Xiuxi Pan of Tokyo Tech.