Streaming Media

Abstract

The color of ocean water provides us tremendous insights regarding the properties of the organic and inorganic particles in it. That ocean water has color, however, is precisely what is holding us back from unveiling the colors of everything else in the ocean, i.e., the colors of the ocean flora, fauna, and the unique habitats that host them. Making matters more difficult, the color of the ocean water shows dramatic variation across latitudes and longitudes, and also changes as a function of time. As a result, water occludes features in underwater scenes a different way in each photo, making them look dull, but also preventing us from efficiently applying powerful computer vision and machine learning algorithms on large underwater image datasets. I will show that a fundamental reason for the lack of a robust method to standardize colors in underwater images has been the common use of an atmospheric image formation model for the ocean. Then, based on a (more) physically accurate underwater image formation model we recently proposed and validated, I will introduce the Sea-thru algorithm that successfully removes water from underwater images in a rapid, objective, and repeatable way. Finally, I’ll discuss the biggest challenges we currently face in underwater imaging, and show some of my ongoing work towards enabling consumer cameras to capture ‘real’ colors underwater real time.

Presenter Bio

Derya Akkaynak is an engineer and oceanographer whose research focuses on problems of imaging and vision underwater. She has professional, technical, and scientific diving certifications and has conducted underwater fieldwork from the Bering Sea to Antarctica. Akkaynak is an honoree for the 2019 Blavatnik Awards for Young Scientists in physics & engineering for her post-doctoral research resolving a fundamental problem in underwater computer vision — the reconstruction of lost colors and contrast which lead to the development of the Sea-thru algorithm. She currently works at Harbor Branch Oceanographic Institute in Ft. Pierce, Florida, USA.

Publication Date

2-19-2021

Document Type

Presentation

Share

COinS