Underwater imaging in turbid environments: generation model, analysis, and verification

Abstract

Optical imaging usually shows significant differences in underwater environments with different characteristics. The general underwater imaging model ignores the influences of turbidity on the attenuation of light. For turbid water with different particle types and concentrations, we introduced background light correction and turbidity factors to express the influences. A parametric turbid underwater image generation model based on dark channel prior theory was proposed, in which, the transmission map changes with the turbidity while the image tone changes with the background light correction. Based on experiments, the high similarity between generated images and real images confirmed that our model can fit the imaging in different bodies of water. Thus, the model can be used to generate a greater diversity of samples of turbid underwater images. We also applied the model to turbid underwater image enhancement. The image quality evaluation results indicated that the model can enhance turbid underwater images well.