Model and calibration of underwater stereo vision based on the light field

In an underwater imaging system, a refractive interface is introduced when a camera looks into the water-based environment, resulting in the usually linear rays of light bending and the commonly used pinhole camera model to be invalid. In this paper, a novel method to model and calibrate the underwater stereo system is proposed. Unlike most existing recent works that use vector operations to express light propagation, this paper remodels the underwater imaging system in the view of the light field, which is more organized and concise. Additionally, a forward projection error function, which is more accurate and efficient, is used for non-linear optimization to minimize the reprojection error. The proposed method allows improved accuracy compared with other methods and is evaluated in a set of experiments.