Universal algorithm for water depth refraction correction in through-water stereo remote sensing

•This paper reports on a universal algorithm for water depth refraction correction algorithm in through-water stereo remote sensing.•The algorithm is reasonable and applicable whether or not conjugate image rays from the stereopair are intersecting.•The algorithm is approximate in most cases, but the relative in-depth error caused by the algorithm itself is small.•The algorithm outperforms the existing algorithms in bathymetric accuracy for an area or a stereopair. In through-water stereo remote sensing, the correction of the effect of light refraction on water depth determination is an unsolved problem. The challenge is that no solution exists for apparent positions when the conjugate image rays are non-intersecting. This paper proposes a new water depth refraction correction algorithm for through-water stereo remote sensing. The midpoint of the shortest line segment between two non-intersecting conjugate image rays is used as the apparent position, so that the mathematical relationship between the apparent and actual depths is established under the “no solution condition”. The new algorithm is reasonable and applicable whether or not conjugate image rays are intersecting. Compared with the existing algorithms, the new algorithm can improve bathymetric accuracy in different degrees. The improvement becomes more significant as the off-nadir view angle difference and the depth increase. Results for Area 1 showed that, compared with the three existing algorithms, the new algorithm improved RMSE accuracy by about 22%, 29% and 24% respectively. The new algorithm enables reasonable water depth refraction correction to be implemented in any case (whether conjugate image rays are intersecting or not).