News + Trends

Researchers remove colour cast from underwater photos

David Lee
14.11.2019
Translation: machine translated

Two researchers are using machine learning to remove the blue or green tint from underwater images - with impressive results.

Underwater images have a strong blue or green cast. This is because the wavelengths of blue and green penetrate much deeper than those of red. The problem can be alleviated with artificial lighting and filters, but this requires considerable effort and does not work in every situation.

Source: https://commons.wikimedia.org/wiki/File:Farbabsorption_unter_Wasser.svg
Source: https://commons.wikimedia.org/wiki/File:Farbabsorption_unter_Wasser.svg

Researchers Derya Akkaynak and Tali Treibitz have developed a new method called Sea-Thru to extract the colours from blue-green underwater photos that they would have without water. The results look pretty impressive.

Source: Scientific American
Source: Scientific American

According to Derya Akkaynak, this is a physically accurate correction and not just colour cosmetics, as would be possible in Photoshop. They photographed the same motif many times with slight variations, always including a colour map. This material was used as training for machine learning.

The degree of discolouration depends on the distance. In a complex three-dimensional structure such as a reef, there are countless different distances from the camera and each pixel must be calculated separately.

Cover image: Scientific American

33 people like this article


User Avatar
User Avatar

My interest in IT and writing landed me in tech journalism early on (2000). I want to know how we can use technology without being used. Outside of the office, I’m a keen musician who makes up for lacking talent with excessive enthusiasm.

11 comments

Avatar
later