This Google AI wants to revolutionize mobile photography. And you don’t need many megapixels or large sensors

  • 22

Computational photography is one of the greatest gifts (and in turn, curses) that technology has brought us. Terminals like the first generation Google Pixel showed that the phone can see even more than we doespecially shining in night photography.

Shooting in low light, however, is a big challenge. The sensor will generate more noise than necessary if the luminosity is scarce and here two decisions can be made: respect the noise to preserve the detail, or remove noise (wash out the photo), thus losing detail. This last decision is the most popular on smartphones, and Google is already training an AI to solve this problem.

Google wants noise-free photos, but with detail

Tg Image 1406768042

MultiNerf is the Google project, launched in open source, to solve the problem of night photography. There is still a certain aversion on the part of users to digital noise, and Google algorithms want to fix the problem without loading the detail of the photographs along the way.

Google is based on NeRF (Neural Radiance Fields), a neural network developed, in the first instance, to generate 3D images based on 2D sets. The purpose of building on NeRF is that, from a 3D image the AI ​​has it easier to analyze the information of the image since it can “move” through it.

“We modified NeRF to train directly on linear raw images, preserving the full dynamic range of the scene. By rendering raw output images from the resulting NeRF, we can perform new high dynamic range (HDR) view synthesis tasks. In addition to By changing the point of view of the camera, we can manipulate focus, exposure and tone mapping after the fact.”

The algorithms of this model analyze the raw data of the RAW file, and use artificial intelligence to detect what the information would be if there were no noise in that scene. In other words, they try to preserve the detail of the RAW, even though the noise is subsequently removed.

Although this model is still being investigated, it is an important basis for the future of computational photography. Let’s hope that within not long, the Pixels begin to implement this technologyand that it expands to the rest of the manufacturers.

Computational photography is one of the greatest gifts (and in turn, curses) that technology has brought us. Terminals like the…

Computational photography is one of the greatest gifts (and in turn, curses) that technology has brought us. Terminals like the…

Leave a Reply

Your email address will not be published.