Underwater image conversion method based on multi-layer plane refraction model with light field processing method

Abstract

In order to adopt the conventional visual measurement technology to underwater application, this paper proposes an underwater image conversion method based on a multi-layer plane refraction model, which converts the underwater image into an air image. First, under the multi-layer plane refraction model, the imaging process of the underwater camera is modelled in the form of a four-dimensional light field parameterization, and the direction vector of the air image is calculated using the light field direction information. Then, the perspective projection transformation is used to obtain the corresponding pixel coordinates of the direction vector and the image plane. Finally, the transformed air image is obtained by interpolation method, using the mapping relation of the corresponding image point coordinates before and after the transformation. Simulation and experimental results show that the average error of the corrected image is 0.56 pixel, and the high-precision air image provides strong support for subsequent underwater 3D reconstruction.