Machine learning improves underwater visibility in real time

Press release /

Fraunhofer IGD is developing special image processing methods based on artificial intelligence to instantly improve visibility under water.

© Fraunhofer IGD
Fraunhofer IGD is developing underwater image enhancement methods based on artificial intelligence.

(Rostock/London) New AI-supported image enhancement by Fraunhofer IGD processes underwater video in real time, making underwater activities faster and more effective. Underwater technology providers can license the solution for embedded systems in smart cameras or on underwater vehicles. One other advantage is that this smart image processing system requires little power to operate. At Oceanology International in London, anyone who is interested can find out more about this new technology and discuss it with experts.

Real-time video processing through artificial intelligence

The Fraunhofer Institute for Computer Graphics Research IGD has years of experience in the field of underwater image enhancement. One of the latest developments is the use of artificial intelligence in the form of what are called “autoencoders” and “generative adversarial networks” in enhancing images and video. An artificial neural network automatically detects the parameters of an underwater recording that require enhancement and makes the necessary adjustments, specifically color correction, noise reduction and sharpening. A second network is responsible for training the first by generating artificial training data, since this is sparse in the specific case of underwater recordings. By integrating AI, parameters no longer need to be manually set as before -- the networks continue to learn and grow by themselves and deliver better and better results.