WaterGAN: Unsupervised Generative Network to Enable Real-time Color Correction of Monocular Underwater Images — mentioned by blt

Thu, 09 Jan 2020 02:09:12 GMT

This would be a great application for deep learning. Use the authors' method to generate a lot of uncorrected-corrected pairs. Or, use a graphics engine to render realistic underwater scenes with and without water color. Then use a convolutional neural network to learn to mimic the transformation. Then any photographer can apply the learned filter without a color card or depth information. Edit: already been done: https://arxiv.org/abs/1702.07392

Paper and discussion history. A mention is not an endorsement.

WaterGAN: Unsupervised Generative Network to Enable Real-time Color Correction of Monocular Underwater Images · HackerPapers discussions

WaterGAN: Unsupervised Generative Network to Enable Real-time Color Correction of Monocular Underwater Images — mentioned by blt