Physics-Inspired Synthesized Underwater Image Dataset

arXiv:2404.03998v1 Announce Type: new
Abstract: This paper introduces the physics-inspired synthesized underwater image dataset (PHISWID), a dataset tailored for enhancing underwater image processing through physics-inspired image synthesis. Deep learning approaches to underwater image enhancement typically demand extensive datasets, yet acquiring paired clean and degraded underwater ones poses significant challenges. While several underwater image datasets have been proposed using physics-based synthesis, a publicly accessible collection has been lacking. Additionally, most underwater image synthesis approaches do not intend to reproduce atmospheric scenes, …

abstract arxiv challenges cs.cv dataset datasets deep learning demand eess.iv image image datasets image processing paper physics processing synthesis synthesized through type underwater