A new AI algorithm called Sea-thru is helping to revolutionize the world of underwater photography. Developed by AI engineer and oceanographer Derya Akkaynak, Sea-thru is able to remove the visual distortion caused by water, creating clearer and crisper photographs.
While this may sound like a niche innovation, it actually opens up a whole new world of possibilities for underwater photography. For example, it has been very difficult to capture excellent photographs on coral reefs, as pictures of them taken without artificial lighting are inevitably washed out and bland.
This effect is surprisingly prevalent even in shallow water conditions, which can selectively absorb and scatter light, making underwater photography of uneven quality. It has been particularly difficult to capture the vibrant colour of the ocean world while underwater, but Sea-thru promises to change this for good.
Akkaynak and engineer Tali Treibitz, her postdoctoral adviser at the University of Haifa in Israel, developed the software together, and outlined the process involved in a paper that the pair presented in June at the Institute of Electrical and Electronics Engineers Conference on Computer Vision and Pattern Recognition.
Sea-thru utilizes stored Image analysis features in order to filter out light absolution and scattering that occurs in the ocean atmosphere. The software is capable of effectively removing image distortion on a pixel-by-pixel basis, restoring colour, and reversing the washed out appearance of photographs.
This is a big deal for marine biologists and other scientists interested in the ocean, as it will make it more feasible for them to engage in tasks, such as the counting and classification of the species in underwater images. Indeed, Sea-thru opens up the possibility that coral scientists could use computer vision and machine-learning in the identification process.
There are still some logistical issues that Sea-thru must overcome, though. For example, the software requires distance information in order to work, with the developers taking numerous photographs of the same scene from various angles in order to collect sufficient information for Sea-thru to work its magic.
But this can be overcome due to the approach that ocean scientists, who already catch distance information on a regular basis, via a process that is referred to as photogrammetry. Akkaynak has stated that the program will work perfectly adequately with such photographs, which would seem to circumnavigate some of the logistical issues.
The Sea-thru software has already attracted a lot of praise from coral and marine biologists, with experts acknowledging that gleaning colour from underwater photographs will enable scientists and researchers to garner a great deal more worth from existing datasets.
These new abilities associated with Sea-thru have effectively been trained into the software, which has been developed by using sets of underwater images in order to teach the algorithm. And as Sea-thru continues to mature, it seems as if the program will have a profound impact on the capabilities of ocean scientists, opening up a new world of crystal clear ocean photography.