Seafloor-Invariant Caustics Removal From Underwater Imagery

Mapping the seafloor with underwater imaging cameras is of significant importance for various applications including marine engineering, geology, geomorphology, archaeology, and biology. For shallow waters, among challenges, caustics, i.e., complex physical phenomena resulting from projection light rays being refracted by wavy surface, likely most crucial one. Caustics main factor during campaigns that massively degrades image quality affects severely any 2-D mosaicking or 3-D reconstruction seabed. In this article, we propose a novel method correcting radiometric effects caustics on imagery. Contrary to state-of-the-art, developed can handle seabed riverbed anaglyph, images using real pixel information, thus, improving matching processes. particular, employs deep learning architectures classify pixels “noncaustics” “caustics.” Then, it exploits geometry scene achieve pixelwise correction, transferring appropriate color values between overlapping images. Moreover, fill current gap, have collected, annotated, structured real-world caustic data set, namely, R-CAUSTIC, which openly available. Overall, based experimental results validation, methodology quite promising in both detecting reconstructing their intensity.