WaterGAN: Unsupervised Generative Network to Enable Real-Time Color Correction of Monocular Underwater Images

• Published in: IEEE robotics and automation letters Vol. 3; no. 1; pp. 387 - 394
• Main Authors: Jie Li, Skinner, Katherine A., Eustice, Ryan M., Johnson-Roberson, Matthew
• Format: Journal Article
• Language: English
• Published: Piscataway IEEE 01.01.2018; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Attenuation; Autonomous underwater vehicles; Color imagery; Columnar structure; Datasets; Deep sea environments; Gallium nitride; Generators; Image color analysis; Image resolution; Image restoration; Light attenuation; Machine learning; Marine robotics; Mathematical models; Nonlinear systems; Ocean floor; Parameters; Pipelines; Remotely operated vehicles; Restoration; Source code; Training; visual learning; Water circulation; Wave attenuation

This letter reports on WaterGAN, a generative adversarial network (GAN) for generating realistic underwater images from in-air image and depth pairings in an unsupervised pipeline used for color correction of monocular underwater images. Cameras onboard autonomous and remotely operated vehicles can...